WO2003031920A1 - Automatic music generation method and device and the applications thereof - Google Patents

Abstract

The invention relates to an automatic music generation method and device. The inventive method consists in producing at least one succession of notes at musical positions respectively. Set pointers are generated at certain ones of said musical positions respectively, each set pointer indicating another of said musical positions, in such a way as to enable a non-sequential reading of the succession of notes. The successions of notes are preferably generated in tones close to one another.

Description

AUTOMATIC MUSIC GENERATION METHOD AND DEVICE

AND APPLICATIONS

FIELD OF THE INVENTION

The present invention relates to a method and a device for automatic musical generation. It applies in particular to the broadcasting of music, to the storage of music, to teaching materials, to music on hold in telephone systems, to music broadcasting by telephone, to telephone rings, to electronic games, to toys, to music synthesizers (hardware or software), computers, camcorders, alarm devices, locksmithing, music telecommunication, music generation subject to external physical parameters, robotics and, more generally, sound illustration and musical creation on any digital medium.

STATE OF THE ART

The methods and devices for musical generation currently known use either a library of precomposed musical sequences and assembled automatically in a random order (first case), or a system of automatic generation of a melody from a random or non harmonic sequence ( second case), or a system for automatically generating a melody then harmonized (third case).

In particular (second and third case), international patent application WO-00/17850, filed in the name of René-Louis Baron, discloses a method of composing a "note after note" melody. It is based on the definition, at any location during which a note is likely to be played, of two families of note pitches, the second family having at least one note pitch which is not in the first family. According to a basic principle of the generation of disclosed music, a note whose pitch belongs exclusively to the second family can only be surrounded by notes from the first family. This method offers almost unlimited possibilities of compositions, with desired styles, while allowing to obtain harmonious and original music.

It is however interesting to deliberately seek to get rid of pre-established rules of composition, so as to obtain unexpected arrangements or sequences. There are methods for achieving such results, but the music produced is then generally difficult to appreciate and lacks consistency.

Other composition techniques make it possible to obtain unexpected effects while ensuring a certain coherence, thanks to the use of substantial databases in which various types of musical arrangements and extracts are stored. The generation of music is then based on the use of strict sequence rules, applied to the information contained in the databases and possibly implementing adaptations of the musical extracts selected in the databases. Such methods, however, require considerable storage resources to provide sufficient diversity of composition. In addition, they require data processing operations which can also be large, notably prohibiting generation in real time. PRESENTATION OF THE INVENTION

The present invention relates to a method and a device for composing music that is both coherent and pleasant to listen to, and at the same time surprising and complex. In addition, the method and the device of the invention make possible the production of compositions with a very reduced storage space and very short processing times, allowing generations of music in real time and possibly requiring only material resources ( processing times, memory space, storage space) of the same order as for the request generation technique WO-00/17850 ..

To this end, the invention relates to an automatic musical generation method, in which at least one succession of notes is produced at musical locations respectively.

According to the invention, playing pointers are generated respectively at certain musical locations, each of these playing pointers indicating another of the musical locations, so as to allow a non-sequential reading of the successions of notes.

This method allows you to start from perfectly melodious music, and transform it by introducing effects of rhythm and / or tone breaks. It is possible to maintain complete control in the dosage of these breaks, which allows the introduction of a greater or lesser level of complexity.

Note pointers can be placed a priori in any location of the song (start of measures, start of time, any musical locations), but their use is particularly recommendable at the start of measurements. Thus, we avoid too brutal effects of rhythmic breaks, while producing original music.

In addition, in this implementation with game pointers, it is preferable to generate the successions of notes in tones close to each other. This maintains a relative harmonic coherence of the piece, avoiding breaks in unpleasant tones.

Preferably, provision is made for a sequential reading of the successions of notes after a reading by means of the play pointers. In this way, the impression of overall coherence of the piece is reinforced.

The method of musical generation advantageously comprises:

at least one operation for defining at least one determined length succession of degrees of a range at musical locations to be played, called a pattern,

at least one operation of duplicating each of these patterns in free musical locations so as to form deduced patterns,

- And an operation for determining the heights of notes of these patterns, according to the degrees of the patterns and of properties specific to each of the patterns.

Thus, we use a musical reference scheme, the motif, when composing musical phrases. This reference pattern, or “father pattern”, is used both for translations over time (reproduction in the deduced patterns, or “son patterns”) and in height (because it only specifies the degrees, therefore the deviations from pitch of the pattern notes relative to each other to others). In fact, two son patterns can be assigned separate note heights. The use of patterns reinforces the cohesion of the piece, and thus makes it possible to make transitions due to the play pointers more easily acceptable by the ear.

Preferably, the specific properties used to determine the pitch of notes include for each of the patterns a general harmony and a chord (which are those in progress for the notes of the pattern).

Songs are generally generated made up of sets of note successions, using, for the operations relating to each of these pieces, prints of a combination of parameters. The method then advantageously comprises an operation of pairing each of the combinations of parameters with a unique digital identifier. The matching operation consists in using this identifier as the seed for initialization of a pseudo-random drawing device.

This is a surprising way of keeping a memory of a large number of parameters, possibly several hundreds, by means of a simple digital identifier. We thus take advantage of pseudo-random generators, based on predefined algorithms and intended to be initialized by a given seed. Such generators are commonly implemented in computers.

The invention also relates to an automatic musical generation device capable of producing at least one succession of notes in musical locations respectively. According to the invention, this device comprises means for generating play pointers respectively at certain of the musical locations, each of these play pointers indicating another of the musical locations, so as to allow a non-sequential reading of the successions of notes.

The music generation device is preferably intended to implement any of the modes of implementing the music generation method according to the invention.

The invention also relates to a computer program product, comprising program code instructions for the execution of the operations of the music generation method according to any one of the modes of implementation when this program is executed on a computer. By “computer program product” is meant any support for the program, which may consist either of a hardware support such as for example a disc, cassette or diskette, or of signals, in particular electronic or in the form of waves.

Various products containing the music generation device are also targeted by the invention, such as in particular an electronic chip, an electronic and / or video game, a telephony device, a music broadcaster, or a sensor of physical quantities.

PREFERRED MODALITIES FOR IMPLEMENTING THE INVENTION

The invention calls upon new musical concepts, this is why the classical terminology is supplemented by a semantics appropriate, used in this application. A lexicon of the words used in the preferred embodiments described in the present application is placed at the end of the description and makes it easier for the reader to understand. For the same purpose, we recall certain fundamental operations of automatic musical composition already set out in application WO-00/17850 cited above.

The American notations are used below for the tones (C, F #, Bb ..., with "#" for sharp and "b" for flat).

The musical composition advantageously calls for the vertical analysis of the different musical parameters of each voice of the counterpoint at each position of the song, which makes it possible to obtain an "intelligent" orchestration where the progression of each voice (from the composition of second voice), has a ratio given arbitrarily with one or more previous votes. This principle of composition thus offers an additional dimension compared to the method set out in application WO-00/17850, relating to a single melodic line.

There is no limit to the number of voices. Thus, for an embodiment using four voices, a distinction is made:

- a first voice (or song), composed from a set of harmonies (or passage chords) structured rhythmically, or on the contrary a first structured voice inducing a series of passage chords;

- a second voice (or melody) created according to the structured harmonies and musical parameters of the first voice;

- a third voice, created according to the structured harmonies and musical parameters of the first and second voices; - and a fourth voice, created according to the structured harmonies and musical parameters of the first, second and third voices.

More precisely, the notes are divided at each location into two families, consisting respectively of base notes and passing notes. Notes played simultaneously (that is to say in identical positions) on several voices are then systematically part of the same family. This composition rule is respected whatever the rhythmic cadences considered.

The base notes are preferably defined as those making up a chord in progress at the location in question (passage chord), as well as these same notes duplicated from octave to octave according to the range of the instrument. The passage notes are then made up of the notes duplicated from octave to octave according to the range of the instrument, from the range of general harmony amputated from the tonic, the third and the fifth of the current passage chord. . These definitions will be explained and exemplified later (see in particular the glossary).

In addition, in the case where a musical phrase of one of the voices ends up in a space which must receive a passing grade, due to its alignment with the family of notes of another of the voices, it is preferably added after from the end of this musical phrase a place to play in the family of basic notes. Thus, it is possible to comply with a rule for the end of musical phrases in base notes.

In other application modes, the number of voices is limited to one, two or three voices. Typically, for obtaining very styles "Typed" (eg Rock and Roll, Bossa-nova, Béguine, etc.), only two voices are generated: the song (first voice) and a second voice for the "responses" to the song in certain passages, or for a real counterpoint in other passages of the song. Other voices, if necessary, are pre-recorded manually by one or more musicians in the form of files, whether or not using the MIDI computer protocol. These pre-recordings can be of variable length, from one to four measures for example. They are played in a loop and transposed with calibration of the note pitches according to the musical parameters of the automatic composition.

The music generation process allows you to compose an infinite number of voices, the only limitation being the number of available MIDI channels (16, 32, 64 ...). It makes possible the composition of a very large number of coherent works in counterpoint meeting the rules of harmony and counterpoint, which govern all the operations of generation of the constituent elements of a musical work. To obtain such results, the random draws are preferably limited by means of constraints imposed by the desired style.

According to a first advantageous mode, the method of musical generation successively comprises the operations set out below.

Operation 1-1: initialization

This operation consists in initializing by pseudo-random draw or not, the pseudo-random draw function of the microprocessor used. The value of the initialization seed, which is generally between 1 and 2,147,483,647 in current microprocessors, makes it possible unequivocal identification of the piece of music composed. Indeed, it makes it possible to obtain the same results for the various successive random draws associated respectively with musical parameters of the piece, these draws being able in certain embodiments of the order of several hundred. This operation makes it possible to recall such or such precise composition, the same initialization value always giving the same music.

By way of illustration, "title%" is a variable designating a specific song (or a "precise combination of notes") among the 2,147,483,647 possible combinations of notes. To identify it, before proceeding to the random draws of all the values of all the musical parameters of the computer program, it is necessary to initialize the pseudo-random draw system of the microprocessor by assigning a value given to the variable “title % ": Title% = 8248 value given to the title variable%

RANDOMIZE title% initialization of pseudo-random draws, where RANDOMIZE designates the initialization operation of the pseudo-random draws.

Advantageously, the seed is initialized to a random value between 1 and 2,147,483,647. Thus, we obtain real random draws. With the previous example, we thus have: title% = INT (RND x 2,147,483,647) +1 value drawn at random for the variable title%

RANDOMIZE title% initialization of random draws where INT designates the integer part and RND a random draw comprised between 0 and 1, so that an integer value drawn at random and comprised between 1 and 2,147,483,647 is assigned as%.

Thus, for a specific style and after this initialization, all the so-called “random” draws made during the calculations of a musical composition, give the same results and are therefore in fact predictable.

This pseudo-random draw function notably makes it possible to identify precisely and unambiguously each piece, for a given style. To do this, we store in a table and for this style, all the results of successive random draws carried out following a random initialization (title% = INT (RND x 2.147.483.647) +1), in the form of a succession of values. This succession of values indeed constitutes a unique identifier of any musical work generated by automatic generation.

The automatic composition program thus appears as a different musical database in which each identifiable work is different from all the others. Advantageously, this specificity is exploited by making available to users (telephone ringtones for example) a bank of musical works all different from one another. It is therefore interesting to ensure that as soon as a user chooses a specific work, it becomes unavailable to future other users.

In an improved implementation, the user of an automatic composition program is authorized to access certain musical parameters of a specific piece (therefore identifiable), and to choose the values - by fixing or modifying them. He can thus give a personal touch to the works generated. To this end, an appropriate user interface is created. This allows the user to change and set certain values of musical parameters (for example relating to tempo, transposition ...).

It is then preferably made so that initially, the generation of music is carried out in a fully automatic way by carrying out the random draws of the values of all the required parameters. Secondly, the user imposes the desired values on some of the parameters. This avoids generating a lag in the results of the successive pseudo-random draws of the program, which would occur if the parameters determined by the user were not the subject of the planned draws beforehand, which would require specific software adaptations. The ease of implementation and the speed of execution are thereby increased.

As an example, the user acts on the “tempo” parameter. First, the program calculates a value for this tempo by: tempo% = INT (RND x 5) + 80, which offers five possibilities (integer value between 80 and 84). Then, the value imposed by the user is taken into account as a replacement: IF user tempo% <> 0 tempo% = user tempo% Operation 1-2: alphabetical title

This operation consists in randomly drawing a coherent alphabetical title thanks to random associations of character strings prepared in data (DATA).

For example, we get:

"The way" + "from the sea" or

"The song" + "of the mountains"

Sufficient strings are prepared to prevent two separate musical works from having the same alphabetical title. However, this title is preferably completed by the style number chosen (upstream of the alphabetical title) and by the numerical title of initialization of the prints (title%, downstream). In this way, we obtain an identifier that is both complete and meaningful. For example, such an identifier consists of:

Style No. 35 “The path to the sea” No. 123.654.

Operation 1-3: style of the song

It is a question of determining the style of the piece to compose. Each style has its own constraints to oppose random draws during the various rhythmic and harmonic operations of musical composition.

Operation 1-4: positions

We determine chronologically units of time space or positions, intended to be played or not, on a scale representing the duration of the piece divided into a certain number of measures, themselves subdivided into a certain number of positions to be played or no.

Operation 1-5: lengths of location groups

One or more lengths (in numbers of positions) of respectively one or more so-called groups of rhythmic locations are determined.

Operation 1-6: locations

This operation consists in determining chronologically each of the locations contained in a group of locations. The number of pitches is equal to the maximum possible number of shortest notes, to be generated per time. For example with a 4/4 measure, we define four positions per beat (sixteenth note), ie the four locations "e1", "e2", "e3", "e4". In a variant with a binary and ternary division, twelve positions per time are defined, ie “e1” to “e12”.

Operation 1-7: lengths of moments

We define lengths (in number of measures) of the different moments of the song. Example: ^1st movement = 16 measures, ^2nd movement = 32 measures, etc.

The previous four operations 1-4 to 1-7 are described in PCT application WO-00/17850, to which reference will usefully be made for more information.

Operation 1-8: rhythmic cadences

The rhythmic cadences are generated, that is to say the positions intended to receive the beginnings of notes. During an operation to generate a rhythmic cadence, random draws of note figures (quarter notes, eighth notes, sixteenth notes, etc.) make it possible to obtain a precise style of music, therefore with fewer risks. If, for example, a 4/4 song is drawn each time, the current time can receive, according to the random draw or not and the desired style, one of the following groups of figures:

- only one black,

- two eighth notes, - an eighth note and a successive half-sigh (or vice versa),

- an eighth note triplet,

- a quaver, a half sigh, and a successive quaver (all in triplet),

- a half sigh and two successive eighth notes (all in triplet), - two successive eighth notes and a half sigh (all in triplet),

- two successive sixteenth notes and a half sigh (or vice versa),

- four sixteenth notes,

In the case of a drawing of longer notes (dotted black, white, round, etc.), the program skips the following positions, which must receive the duration of the note.

Depending on the desired style, each group of figures has its own density (number of chances to be chosen). In the first voice in “Fugue” style, for example, the group “four sixteenth notes per beat” has the highest density.

For this operation of generating rhythmic cadences, the procedure is preferably as follows. During an operation for generating a "mother rhythmic cadence", a random length of cadence is given, the unit of length here being time. Each cadence obtained from "n" time is then duplicated later in the song to free places in the song and chosen randomly. This duplication is carried out or not a certain number of times, also obtained by random drawing.

According to a second advantageous mode, degrees of pattern are randomly generated for each position to be played of each mother rhythmic cadence. The set of pattern degrees obtained in the same mother rhythmic rhythm constitutes a father pattern. This father motif is then duplicated in the same way (in the same positions) as the mother cadences were during the rhythm cadences operation. These duplications are called child patterns.

According to a third advantageous mode, applicable to the generation of note patterns, a song is produced in the same musical sentence, either ascending or descending, or alternately and randomly both. “Ascending” and “descending” mean respectively a succession of notes from low to high and vice versa.

In the case of a rising voice song, for example: a / we choose the first pitch of the musical phrase to be composed randomly from among the lowest notes of the current range (first third for example), while respecting the note family of the playing position concerned (this is valid for all the notes of the musical phrase to be composed); b / we take the following notes in the current scale, one or two degrees higher than the previous one; c / in the case where the highest possible degree is reached (according to the available range), the current note is considered as a first note of musical phrase (a /). In another application mode, and still according to the example above, the number of positions to be played (in the musical phrase to be composed) is calculated, and the first note is chosen from among those sufficiently low in the current range , to allow a total ascendancy of all the notes of the musical phrase.

According to a fourth advantageous mode, crossover of note patterns between the different voices is generated. For example, if the first voice is mainly ascending, the second voice is mainly descending and so on.

This method of obtaining ascending or descending movements of the different voices makes it possible, in another mode of application, for example more precisely concerning video games, to affiliate with each voice:

- a different sound program (violins, brass, etc.);

- and / or a specific character or team (the players): according to the evolution of the game, the singing notes of the voice associated with the winner go up the range while the voice associated with the loser undergoes a downgrade.

According to a fifth advantageous mode, game pointers are generated at randomly chosen positions of the piece. Each of these pointers indicates a position of the song other than that where it is placed. One can thus, when all the operations of musical composition of a piece are finished, read Qouer) this piece in a non-chronological way. To obtain a rhythmically coherent music, that is to say respecting the signature (s) of the piece (3/4, 6/8, etc.), we choose the positions of the play pointers among the first positions of the different measures of the total piece. In variants, the positioning of the pointers is authorized at the start of times, or even moments of the piece, which reinforces the impression of breaks.

The breaks introduced enrich the possibilities of composition and the complexity of the songs, and can thus increase the entertainment and the pleasure of the listener.

Preferably, we choose neighboring tones throughout the song. In this way, despite the breaks in the reading, the whole remains harmonious to hear. In addition, the reading of the piece is advantageously completed in the order indicated by the play pointers, by a complete reading in the sequential order. This reinforces the impression of consistency and aesthetics. The number of game pointers used depends on the breaking effects that one wishes to produce: the more pointers there are, the more the music becomes rich and complex, but also more difficult to grasp. An interesting possibility is to place a pointer at the start of each measurement. Another is to place only a limited number of pointers (for example one every ten measures), but in any location.

According to a sixth advantageous mode, two families of note pitches are used at any time: - a "basic family", whose so-called "basic" notes are those of the current chord (perfect or not), duplicated from octave to octave over the whole of a defined range; - and a “passage family” whose so-called “passage family” notes are those of the current general harmony scale, reduced by:

• the tonic, "the third (minor or major as the case may be),

• and the fifth (fair, reduced or increased as the case may be), of the perfect or enriched associated passage chord (encrypted or compound) in progress.

These degrees to extract (tonic, third, fifth) are not the same as those of the range of the current general harmony (or "tone" indicated by the key or by accidental alterations), but those of a new range resulting from the latter whose tonic (or 1 ^st degree) has for note the name of the chord itself.

The additional notes of an enriched chord, i.e. the seventh minor or major, the sixth, the fourth, the ninth, etc., belong to both the basic family and the passing family.

When one of the additional notes of an enriched chord is of the family of passage notes, it prevails over the corresponding note of the general harmony, which has the effect in some cases, of the modification of this note in the range of general harmony. So for example, the 7 ^th degree of a range of general harmony in C originally in harmonic mode, that is to say here the note "if", becomes the 7 ^th degree of a melodic range of Do ie "si-b".

As an illustration for the Cm9 chord and the general harmonic C minor: - the Cm9 chord consists of the notes: C, E-B, G, D;

- only the notes: do, mi-b and sol are removed from the general harmonic to define the passage family;

- therefore the note D (second degree of the scale) belongs to both the basic family and the passing family.

To obtain “simple” music in the melody, the families of note pitches are advantageously determined by associating the families with locations of notes to be played in each measure. Thus, the location “e1” is a so-called basic location, that is to say that must receive a note taken from the so-called basic family. According to a particular implementation, the desired style of music (simple in its melody) requires obtaining, most often, the eighth note as the shortest note length, the sixteenth note being tolerated only as accident, therefore in low density. The locations "e1" and "e3" then have a normal density (medium to high) and the locations "e2" and "e4" have a very low density. In this case, the locations "e1" are locations to receive base grades and the locations "e3" are locations to receive pass marks.

Detailed methods of implementation for this use of the locations can be found in patent application WO-00/17850.

To obtain a more complex or richer music, families of note pitches are advantageously determined by applying a method of alternating families, and while respecting the methods described below. We observe the alternation of base notes and passing notes within the same musical phrase, but with the possibility of obtaining, depending on the desired music style, two base notes (or more) which following. Only successions of passing marks are prohibited, except in the case of the 6 ^th and 7 ^th degrees of a diatonic range of any mode and where these two degrees are immediately preceded and followed by degrees from the family. basic.

An example is given in table 1, for a 4/4 measure with a diatonic range of C on a perfect chord of passage C. The numbers following the note values identify the octaves in ascending order (so do4 denotes the note "do" of the fourth octave).

Table 1

Another example shown in Table 2 concerns a tone (general harmony) of C Major, on a single 4/4 measure:

Table 2

A third example, referring to Table 3, covers a series of notes on an agreement being ^7th Do Maggiore (CM7) in C Major tone and one in 4/4:

Table 3

The first note (si3) is included in those which constitute the CM7 passage chord (do, mi, sol, si), so that it can be according to its location in the current musical phrase, i.e. base note , or passing grade.

The octave positions of the notes in the tessitura of the instrument used do not affect the membership of these notes in a basic or passing family. The same is true, more generally, of the degrees of the current range, belonging to one or other of the families.

For example, an associated chord being C (C Major), we have the following basic notes and degrees:

- do1, do2 ... doδ (degree 1); - millet, mi2 ... mi8 (degree 2);

- sol1, sol2 ... solδ (degreeδ) and for notes and passage degrees:

- re1, re2 ... reδ (degree 1);

- fa1, fa2 ... fa8 (degree 2); - Ia1, la2 ... la8 (degree5) - si1, si2 ... si8 (degree5)

It is important to clarify that these are the degrees of a diatonic range (degrees 1 to 7) in Ionian harmonic mode and not of pattern degrees (degrees 1 to 6). In other embodiments, the diatonic range is respectively of Dorian, Phrygian, Lydian, Myxolidian, Aeolian, Locrian, or Melodic modes and the operation is the same for each of them since they are degrees of the scale and not of the notes themselves. The pentatonic scales do not escape the rule of note families although the base notes are more numerous. An example is given in Table 4, where B / P means "base or passage".

Table 4

We note that the 2 ^nd and 5 ^th degrees can be either base or passage.

The family of the first and last notes of a musical phrase should generally be basic. An exception for the first note is possible in cases where the alternation process is respected and the next playing note, therefore the second of the musical phrase, is not far from its attack (beginning of note) by more than '' a given maximum distance away. This distance is, for example, half a beat for an average tempo and a non-wide melody. It is generally mobile and proportional to the tempo: the slower the tempo and the more we reduce this maximum distance (and vice versa). In the case of so-called wide songs, this distance can be greater. Advantageously, it is defined as a fraction of the distance which separates the attack of the last note of a musical phrase from the attack of the first note of the following phrase, and preferably as equal to this distance (for example: a measurement). This prevents all the notes from being basic notes in a wide song.

In an example shown in Table 5, a musical phrase includes two 4/4 measures on a C major chord:

Table 5

In this alternation methodology, the operation of assigning families to the different locations ("e1", "e2" ...) of time is not necessary.

According to a seventh advantageous mode, the generation of the notes is based on the successive operations described below.

Operation 2-1: family of notes for beginning of sentences

Families of notes to be played are defined at the positions to be played, corresponding respectively to the beginnings of musical phrases. These beginnings of sentences include on the one hand the first position to play the piece and on the other hand the other positions to play each located at a significant distance from the previous position to play (which constitutes the end of the previous musical phrase). This distance is always greater than those which separate the playing positions from the musical phrase, and allows for example breathing for a trumpet player.

Operation 2-2: note families for playing positions

We define a note family (base or passage) for each position to be played obtained by an operation of generation of rhythmic cadences and by the option taken concerning the assignment of families to the positions to be played: alternation method or method "by locations ”(“ e1 ”,“ e2 ”...).

Reference will usefully be made to application WO-00/17850 to obtain more detailed information on the two previous operations 2-1 and 2-2.

Operation 2-3: filtering for alternating families

This filtering operation prohibits, insofar as the alternation method is used for the determination of note families, any succession of passing notes. An exception is made for the notes corresponding to the 6 ^th and 7 ^th degrees of the current scale, which alone can follow one another in ascending or descending order provided that this group of two notes is included in the same musical phrase and that 'it is surrounded by basic family notes. Operation 2-4: filtering out duplicates

We regulate the notes called "duplicates", by authorizing or not the successions of base notes of the same pitch.

Operation 2-5: filtering differences in note pitches

The succession of notes whose interval is greater than a predefined maximum interval is prohibited.

Operation 2-6: end of sentence notes

Base notes are generated at the positions corresponding to the musical sentence ends defined above.

The previous three operations 2-4 to 2-6 are described in part in PCT application WO 00/17850.

Thanks to each of these arrangements, the music generated is coherent, and pleasant to hear, since the musical parameters are linked together by constraints. In addition, musical generation is easy and can, depending on the application of the device, be done without any human assistance.

APPLICATIONS

The invention also relates in particular to the following products, comprising a device according to the invention: a compact disc, an information medium, a smart card, a modem, a computer and its peripherals, alarm, toy, electronic game, electronic gadget, postcard, music box, camcorder, image recorder, sound recorder, electronic music card, synthesizer (hardware or software ), a music broadcaster, a music telephone server, a music telematics server, a music generator, an educational book, an educational digital data medium, a work of art, a radio transmitter, a television transmitter, a television receiver, a cassette or audio disc player and player, a cassette or video disc player and player, a telephone, an answering machine, a switchboard, a robot, a physical size and a lock system.

The invention also applies to: - a means of storing information (readable by a computer or a microprocessor) keeping instructions of a computer program and allowing, locally or remotely, the implementation of the process invention; this storage means may or may not be partially or completely removable; a means of access to a particular composition obtained by implementing the method of the invention;

- And a means of comparison of musical data available with reference compositions obtained by means of the method of the invention.

These various devices and means directly benefit from the advantages of the method and the device of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

Other advantages and characteristics of the invention will emerge from the following examples of embodiment and implementation, which are in no way limitative, with reference to the appended figures in which:

- Figure 1 shows a first part of a flowchart showing steps of the method according to the invention;

- Figure 2 shows a second part of the flow diagram of Figure 1;

- Figure 3 shows a flow diagram of different stages of the process using degree patterns;

- Figure 4 shows schematically in the form of a functional diagram a first and a third part of a musical composition set capable of implementing the method of the invention, these two parts respectively concerning a server and the connection between this server and a user; and

- Figure 5 shows schematically a second part of the assembly of Figure 4, this second part concerning the user.

EXAMPLES OF IMPLEMENTATION AND IMPLEMENTATION

In modes of implementation of a method of generating music described below, a Web site maker is allowed to illustrate musically, very quickly, elements of text, images, animations and movies (web pages). We use a set of musical composition divided into three parts (Figures 4 and 5): a first part 1 owned by a server administrator or "Webmaster", a second part 2 installed at a user's home, and a third part 3 connecting the first two parts 1 and 2 to each other.

The first part 1 of the set (Figure 4) includes music composition functionalities incorporated into a computer program of a server in hardware or software form, provided with a user interface for the webmaster. As a variant, these functionalities are implemented by an autonomous automatic composition device connected to the server.

Part 1 includes a computer 4 provided with a keyboard 407 and a screen 408. The computer 4 includes:

- a read only memory (ROM) 400 (Read-Only Memory) 400, intended to contain in particular the computer program and the user interface;

- a random access memory or RAM (for Random-Access Memory) 404, designed to contain tables of game variables and game data;

- a clock 402; - an input-output port 403;

- and a central unit 406.

The composition program makes it possible to select a style of music from different styles (for example 50) displayed on the computer screen 408, then to quickly and successively listen to several musics or beginnings of musics, in the chosen style. In practice, it proves possible to compose a hundred different musics in less than a second with a simple usual personal computer. The user interface also allows the webmaster to personalize the musical works generated by the automatic composition program. It can thus in particular modify, in the RAM memory 404, values of purely musical parameters (transposition, tempo ...) and of the instruments of each part of the orchestra, or carry out a personal mixing of these orchestral parts (volumes, pan, effects ...).

The webmaster can keep in a storage space (any digital medium such as for example a hard disk, a floppy disk or a cassette) identifiers of composed music. The program then simply records a seed for initialization of the random draws (between 1 and 2,147,483,647), the style number chosen (for example from 1 to 50) and possibly different values of musical personalization parameters. Such a recording makes it possible to create only a very small data file (barely a few bytes), while allowing the representation of precise music that can last several minutes.

To re-listen to one of the music referenced in such a file, it is enough to load the latter, and the musical composition program then reconstructs exactly the selected piece of music in a very fast time (for example of the order of 1/100 ^e of a second with a usual personal computer). This provides a particularly efficient form of algorithmic compression of musical data: while an average MIDI file containing 3 minutes of music usually weighs around 40 to 60 kilobytes, the above recording allows storage of 50 bytes (almost a thousand times less). The second part 2 of the assembly (Figure 5) includes a computer 5 of a surfer, equipped with a monitor 508, as well as a MIDI interface 513 connected to amplifiers and speakers 512 via a card sounds 509 and an audio output 511. The computer 5 includes a ROM memory 505 and a RAM memory 504, a clock 502, an input-output port 503 and a central unit 506. The ROM memories 505 and RAM 504 are respectively designed to contain information similar to that of the ROM 405 and RAM 404 memories of the webmaster's computer 4 for generating music, but without a user interface.

The program is installed as a plug-in or "plug-in" on the user's hard drive. It is also authorized to be launched only by an order coming from outside, for example by a predefined path on

Internet. Once launched, the program calculates a musical work in a given style and possibly reconstructs a specific work.

The third part 3 of the composition set (Figure 4) connects the first and second parts 1 and 2. It comprises means of data transfer including a hosting website 409, a modem 413 on the server side, a line telephone 412 and a modem 414 on the Internet side. The data transferred relate to coded commands for triggering the user program and to attributes of calculations (style, numerical title, values of the personalization parameters). These orders accompany the web page designed and illustrated musically by the webmaster, all of which is available on the website of the user, client of the webmaster. The transfer is therefore made from this website to the user's 414 modem. The transfer of the above orders and web pages can be carried out on an Internet or Ethernet network, in particular via telephone lines, radio links, power lines, cable, or any other digital transfer medium.

The first part 1 of the composition set - concerning the webmaster's composition program - is intended to implement the operations described below, with reference to Figures 1 to 3.

These operations make it possible to obtain a very large number of musical compositions in counterpoint to four voices, each of the styles and each of the compositions in this style being callable by simple initialization of the seed of random draw of the microprocessor used (initialization step 101). Each composition therefore has here two reference values (seed and style) and an alphanumeric title which allows its identification:

the value of the initialization seed (initialization step 101) of each style can be between 1 and 2,147,483,647;

- the value of the style depends on the number of styles (eg: 1 to 50); - And the value of the alphabetical title (draw step 102) here represents a maximum of 30 characters (ASCII), the alphanumeric title combining this alphabetical title with the two values above.

Here, the song is first composed and then played. In another embodiment, the game is performed as the values of the different musical parameters of the composition are obtained. However, to benefit from the advantages offered by the advance draw, the composition is then carried out in groups of several positions in order to know at any time of the composition the near future of the note (s) to be processed. One will find a description of anticipation of drawing usable for this purpose in patent application WO-00/17850.

In the embodiment described and shown, the successive operations of musical composition are as follows:

A / Style of the song

It is determined by means of a random drawing operation (step 103) or by selecting an orchestra or style, from a set of orchestras composed of instruments specific to a specific style of music (classical, variety, jazz, ...). This information is accompanied by the proper values for the following entities:

- fixed or random adjustment parameters for each instrument or for each part (voice) of the same instrument (general volume, volume of the instrument or of the voice, reverberation, echoes, pan, envelope, brightness, tune, etc.);

- value or range of values of speed of execution (tempo) of the game whose result influences the following operations, the melody and the musical arrangement being able to be denser (more notes) if the tempo is slow, and vice versa;

- transposition parameter, which constitutes a tone parameter (draw step 104);

- number of possible places to play per time, that is to say the choice of the general rhythmic style of the piece to be composed (example:

12 places per beat for a ternary and / or binary rhythm which allows to obtain eighth triplets, sixteenth notes sixteenths, etc., or even only 4 places per time for a rhythm very simple binary whose shortest note is the sixteenth note, here called "e1", "e2", "e3" and "e4");

- number of beats per measure, here 4, that is a 4/4 measure, offering 16 (4x4) positions of notes or rests per measure; - general structure of the piece (random definition or not), that is to say lengths of the moments (in number of measures), number of repeats, locations of sending and resuming;

- family of rhythmic cadences of the first voice;

- family of rhythmic cadences of the three other voices, of accompaniment, in order to obtain a variety of arrangements in the same style, this choice determining later the dynamics of each location in the calculation of the velocity (volume of each note) ;

- type of scale to use (varieties, classical, blues, jazz, Chinese, Japanese, etc.); and - family of passage chords (perfect, composed, blues, Afro-Cuban, Spanish, Slavic, oriental, jazz, etc.).

B / Main tone of the song (or basic general harmony)

We perform a random draw of the tone (step 104) among 11 tones:

C, C # (Db), D, D # (Eb), E, F, F # (Gb), G, G # (Ab), A, A # (Bb) and B.

We will see later that a harmonic modulation operation is allowed thanks to markers (draw step 107) generated by the random draw of the chords of passage and thanks to the fact that the sequences of chords of passage of the piece are made up exclusively chords of tones close to the basic general tone (C).

C / Main harmonic mode of the song

It is determined by a random draw operation (step 105), among two modes here: major harmonic and minor harmonic.

D / Song passage chords

Passage chords are determined (step 106) corresponding to the chords of neighboring tones of the current general harmony. These chords are selected from the minor and major chords of tones close to the basic key of C (C, Am, F, Dm, G, Em).

Passage sub-agreements are also determined, making it possible to pass from one passage agreement to another according to the rules of the art. These are dominant seventh chords, diminished, eleventh, fourth ... according to the desired style, which allow directly or indirectly (that is to say with one or more chords respectively) to access a chord from your neighbor (E7 to access Am, D7 to G, Fdim to Am, Dm11 via G7 to C, etc.).

There is also an operation of random choice or not of density of passage chords within one or more measures. If the density chosen is the lowest, the song contains only one “passage” chord. Otherwise, the higher density allows a change of passage chord every time (four chords per measure at 4/4). A first operation of randomly drawing chords is made from available chords in the family of chords defined by the style chosen above, for each measure, according to the constraints of good practice.

An example is given in table 6 with 2 measures at 4/4:

Table 6

A second operation of randomly drawing chords, constituting passage sub-chords, is made among chords available in the chord family defined by the style chosen above. These sub-agreements are used for a possible insertion of agreements within each measure, according to the constraints of good practice. Here, the insertion is on the ^3rd beat of the measure, as shown in Table 7 extending the example above (Table 6):

Table 7

We then perform a random draw operation of pre-programmed chord sequences (always step 106) in the chord family defined by the style chosen above and intended for each of the moments: 1 ^st , 2 ^nd movement then 3 ^rd movement. The determination of the passage and passage sub-chords of these preprogrammed chord suites is the same as described above.

The choice of one or the other of the preceding drawing operations is itself arbitrary or random.

The embodiment described and represented generates “classic” style music. Thus, the available agreements are minor and major with or without: - fourth (or eleventh),

- diminished fifth,

- increased fifth,

- addition of sixth,

- addition of dominant seventh, - addition of major seventh,

- addition of ninth (or major second),

- or reduced.

An operation is also carried out for determining, for each position of the piece, or more generally for a given number of positions, the type of truncated scale used, that is to say a range passing exclusively through one of the two degrees of next range: sixth degree or seventh degree.

El Harmonic Modulations of the Song

One or more markers are determined (step 107) allowing harmonic modulation. According to different implementation methods, which can be combined, they consist of precise locations in the song (for example the beginning of the ^2nd movement) or in a condition of concatenation of precise passage chords (for example two C chords which follow each other on two consecutive measures).

The operation of generating harmonic modulation markers determines the position or positions from which or from which the general harmony or tone can change. This change is arbitrarily defined according to the desired style. Choosing for initial passage chords the chords and sub-chords of neighboring C tones, in addition to the general harmony chord C, allows harmonic modulation by simply transposing the chords drawn in the current song . Such a transposition is carried out each time that the modulation is imposed by one of the harmonic modulation markers. In addition, the tone or a certain length of the chord preceding the modulation is replaced by a transition chord (preferably the dominant 7 ^th chord: C7 for F or Fm, Bb7 for Eb or Ebm). Thus, this modulation does not occur suddenly.

For example, if the presence of markers is conditioned by the succession of two identical chords and if the random drawing of the passage chords means that two Cm chords of one measure each follow each other, the first becomes major with the addition of 7 ^th minor and the position corresponding to the start of the second chord receives a tone change marker. This new tone is chosen randomly or not from those corresponding to the rules of harmonic sequences (here: Fm or F) and is obtained by simple transposition of the current tone. The return to the initial tone is done using the same process by means of the dominant 7 ^th transition chord (here G7, which leads to C or Cm). It is however possible in certain styles of music to transit from one tone to another without a passage agreement (suddenly without warning). In this case, only the marker is used without creating the passage agreement.

As a variant, the introduction of the harmonic modulation markers at the determined locations is subject to a probability of occurrence (for example one chance in two).

In another example of harmonic modulation, the condition for changing the tone consists of the succession of two chords of the general harmony on two consecutive measures, and a given probability of doubling any measure which is associated with the chord d is applied. '' general harmony (for example one chance in six). This increases the rate of modulation.

F / 1 ^st VOICE (or main melody)

F1 / rhythm of the voices 1 ^st rate (step 108)

A first operation consists in determining a length of play for each voice. Here, the different lengths of rhythmic cadences of the voices are valued so that a / the 1 ^st and 2 ^èmβ voice come from the ^1st movement: two votes; b / the ^3rd voice between that from the ^2nd motion: three voices; and c / the 4 ^{th e} voice between that from the 3 ^rd movement: four voice

(tutti); which creates a growing amplitude from the beginning to the end of the song. A next operation consists in generating rhythmic cadences (mother cadences) here of variable and random lengths (from a single beat to several beats, in one or more measures: drawing step 112), reproduced or not a random number of times or not at random locations or not but blank throughout the song (cadence (s) girl (s)). This operation (drawing step 113) determines the positions to receive a figure of note to play and more precisely the positions of a start of note to play ("notes-on"). The other positions must receive rests, note dressings or end-of-note orders (“notes-off”).

A specific density coefficient is applied to the note figures available for each movement of the song and is valued according to the desired style of music. For example, the first movement has only quarter notes, quarter notes and circles, the other note figures therefore having a density equal to zero at this time. The density of the round can be 25%, that is to say that there is a one in four chance of obtaining during the random drawing of the figures of notes, a round starting on the position of this drawing.

The density of the rhythmic cadence is preferably inversely proportional to the speed of execution of the piece. Thus, the faster the piece, the lower the value of this density coefficient.

A next operation consists in determining for each movement (randomly or not), lengths), frequency (s) and location (s) of ventilation of the rhythmic rhythm of the song calculated above. This operation allows a "breathing" at regular intervals or not. For example, in the ^1st movement, 7 positions every two steps from the ^6th position of the 1 measures ^era are positions not to play (silences).

This aeration operation is carried out:

- either for each voice independently of the other voices;

- either globally on all voices when the different rhythmic cadences (here four voices) are calculated; thus, the four voices "breathe" at the same time;

- either in two ways.

F2 / Times notes between the attacks of the ^1st vote

The durations are determined, expressed in numbers of positions separating the future notes to be played, also taking into account the duration between the start of the song and the first note to be played and the duration between the last note to play and the end of the piece. These last two durations can in some cases be equal to zero. We will see later that each note duration is recalculated and personalized or left in the Qeu legato state).

F3 / Two families of notes

Two “note families” are generated, the basic notes and the passing notes, as defined above and stated in the glossary. This distinction of two families is also described in patent application WO-00/17850, which discloses some exemplary embodiments.

For example, the general harmony being C and the passage chord

(your neighbor of C) being F, the base notes are the notes below underlined and making up the passage chord F, and form the family of base notes: la, si, do, ré, mi, fa, sol, la, si, do, ré, mi, fa, ... The other notes form the family of passing notes (removal of the tonic, the third and the fifth, i.e. of all the notes of the Fa agreement).

In the embodiment described and shown, and with the exceptions described below, the melody consists of alternating passing notes and base notes.

F4 / Family positions to play ^1st vote

The family of notes that each position to be played must be determined (step 114):

- according to its position in the current musical phrase;

- according to its past and future neighborhood, that is to say the duration values which separate it from the previous playing position and the following playing position;

- possibly according to its rhythmic location in the current measure; and

- according to the family of the previous note. In the embodiment described and represented, and with the exceptions described below, the melody is constituted by means of a technique of alternating passing notes and base notes in the same musical phrase, and not by a determination family according to location to play (such an embodiment, also applicable to the present invention, is explained above and is detailed for certain embodiments in PCT application WO-00/17850).

The exceptions to the alternation of base / passage notes are indicated below.

Exception 1

According to the desired style, several base notes, identical or not, follow one another.

Exception 2

The last note position to play a musical phrase receives a note from the so-called basic family, regardless of its location in the current measure. Here, an end of sentence position is considered as such if it is followed by a minimum of three positions not to play. This minimum deviation can reach the quarter note and its value is inversely proportional to the tempo of the song. Indeed, on a very slow tempo, it would be unsightly to hold a passing note for too long. The solution consists in making a harmonic correction by changing the passage chord so that the same note (eg below: re3) passes from the passage family to the basic one, or keeping the associated chord but in choosing another note, this time in the base family. An example with very slow tempo (50 to quarter note) is given in its unsatisfactory form in table 8:

Table 8

A first solution, with harmonic correction, is indicated in table 9:

Table 9

and a second solution, with change of note family, is indicated in table 10:

Table 10

Exception 3

The first note position to play a musical phrase receives a note from the so-called basic family regardless of its location in the current measure, except in rare arbitrary exceptions (for example in jazz styles) but provided that then the alternation of note families.

Exception 4

A style is given by systematics or "ticks". Here, for example, in the case where the end of the musical phrase overlaps two measures and the last chord of passage of the first of these two measures is different from the first chord of the second, the last note of the first measure is drawn systematically in the family of base notes.

Exception 5

In the embodiment described, the 6 ^th and 7 ^th degrees of a diatonic scale are the only notes of the passing family that can succeed one another, in ascending or descending order, provided that this group of two notes is surrounded, imperatively in the same musical phrase, by two notes of the basic family. However, initially, for reasons of speed and ease of calculation, only one of these two degrees is available in the diatonic range used when drawing the degrees of patterns (see § F5). The other degree is added or not during a subsequent operation. We will see in the example of the following operation that this system makes it possible to obtain a succession of degrees alternating chronologically the two families (base and passage).

F5 / Degrees pattern and reasons for the ^1st vote

A random operation is carried out to determine the degrees of pattern for each position to be played of each “mother” cadence (step 115). The set of pattern degrees obtained in the same mother rhythmic rhythm constitutes a father pattern. This father motif is then duplicated (step 116) in the same way (at the same positions) as the mother cadences were during the operation of the rhythmic cadences (cf. § D1). These duplications are called here son patterns.

The degrees of the pattern should not be confused with the degrees of a diatonic scale or other ranges of different harmonic mode.

A general example with a rhythmic cadence of a measure at 4/4 is exposed in table 11, for a piece in C major with a passage chord "C": Table 11

Here, for reasons of speed of computer processing, the degrees used are the degrees of the current diatonic range amputated randomly or arbitrarily from the 6 ^th degree or the 7 ^th degree (both being of the so-called passing family), which makes it possible, when drawing the note pitches, to obtain the base notes at the positions of the odd pattern degrees and the passing notes at the positions of the even pattern degrees.

The drawn degree (6 ^th or 7 ^th ) is stored in the same position in a special table called "six-or-seventh" which is later used to generate the melody and additional voices in order to avoid dissonance of seconds minor or major unwanted, knowing that here a song is made up of four different voices.

The available degree (6 ^th or 7 ^th degree of the diatonic scale) can then be randomly inserted or not in the melody at the end of the next operation of the drawing of the pitches of notes, on condition of keeping a trace of it in the table of note families of playing positions mentioned above.

Two examples of valuation of pattern degrees in a diatonic range of C Major are respectively: do re mi fa sol la do

1 2 3 4 5 6 7 and: do ré mi fa sol si do 1 2 3 4 5 6 7

The pattern degrees have values between 1 and n, the latter being the highest value accepted according to the range of the instrument called to play the voice during composition, the value "1" corresponding to the lower.

During the random drawing of the degrees of pattern, a first filter makes it possible to control the interval (in semitones) which separates this one from the previous one (except for the first drawing). The value of this interval is defined by the desired style. If the interval is too large, a new draw is made until an interval accepted by the filter is obtained.

A second filter allows you to control duplicate pattern degrees. A repetition of a degree of pattern on the same passage chord indeed generates one or more repetitions of the same pitch in the melody. Generally, and depending on the style of music to be composed, this second filter allows duplicate pitch pitches or not. If duplicates are accepted, it lets pass a predetermined number of duplicates (1, 2, 3, etc.) depending on the style. When this ceiling number is reached, a new draw is made to replace the repeated degree, until a different degree height is obtained.

The interval and duplicate filters can be applied to the entire song or only within the same musical phrase. In an implementation variant, these filters are applied to the note-taking operation, as described in application WO-00/17850. In this case, the interval and duplicate filters directly control the pitch of the notes, not the pattern degrees.

At this stage of the composition, it is not yet a question of note pitches according to a given harmony and we will see below that two father and son motifs, therefore identical, can receive different note pitches.

An example of a random drawing of pattern degrees at the playing positions of a "mother" cadence for a range of a single octave (do3 to do4) is set out below. Chronologically, and from low to high, the available pattern degrees are initially valued from 1 to 8. The pattern degree "1" corresponds to the tonic (do3), the degree "2" to the second ( d3), degree "3" to the third (mi3), etc.

Thus, the following random pattern of a "mother" rhythmic cadence:

"1 - 3 - 2 - 5 - 7" induces the following notes:

"Do - mi - ré - sol - do" in the diatonic scale of C Major.

If the mother rhythmic cadence is duplicated a certain number of times during the previous operation of generating rhythmic cadences, the pattern of this mother cadence is duplicated under the same conditions at the playing positions of the daughter rhythmic cadence (s). A next operation consists in generating crossovers of note patterns between the different voices (for example: first mainly ascending voice and second mainly descending voice, and so on).

F6 / Heights notes of the ^1st Voice (General)

A note pitch drawing operation (step 117) for each position to be played takes into account:

- the degree of pattern at this position including the indication of note family (accepted value of the degree of pattern: odd = Basic family, even = Passage family);

- the reading of the table called "sixth-or-seventh" in the case where the value of the degree of pattern is equal to 6;

- general harmony (or tone);

- the passage agreement (perfect or enriched) in progress;

- and the available range (here diatonic) duplicated from octave to octave according to the range of the target instrument.

If the note pitch to be processed is included in a son pattern, the note pitch draw also takes into account the possible presence, in the note family of this playing position, of the note pitch already drawn at the same relative position in the father motif.

F7 / intervals between the notes of the ^1st vote

For each note pitch draw, a pitch comparison operation with the previous note is performed. If the interval between the two compared notes is greater than the maximum authorized interval (for example: five semitones), we proceed to a new drawing operation of note pitch at the current position in the same family of notes.

F8 / Duplicates of successive notes of the ^1st vote

For each note pitch draw, a pitch comparison operation with the previous note is performed. If the note drawn has the same pitch as the previous note, we proceed or not, according to a tolerance coefficient of duplicates of notes, to a new operation of drawing of note pitch at the current position in the same family of notes.

F9 / Velocities notes of 1 ^èrβ voice

A semi-random generation operation of note velocities of the ^1st voice, depending on their location in time and their positions in the song, is implemented.

F10 / Playing lengths of the notes or duration of the notes of the 1st voice

Then an operation is carried out, randomly or not, of “note-off” positions. These determine, for each position, the end of the length of the previous note. In this case, the shortest length is the sixteenth note (only one position of the piece) and the greatest possible length is that separating the previous note from the note following it or, for the last note, the length separating the latter from the end of the song.

G / 2 ^nd , 3 ^rd and 4 ^th voice (steps 109, 1 10 and 1 1 1)

G1 / Rhythmic cadences of voices 2, 3 and 4

A first operation consists in generating a rhythmic cadence so as to obtain a response to the song. This applies here only to the ^2nd voice that answers the voice 1 ^era. It is operational from the time the length of a note of 1 vote ^era is long enough. This minimum length is arbitrarily chosen according to the desired style (eg three full beats), but it is also proportional to the speed of the song's performance: the faster the song is, the greater this length. In addition, for the rest of the song, the rhythmic cadence generation operation for the 1 ^st voice (see § F1) applies similarly to the 2 ^nd voice.

A second possible operation (one chance in two) is based on a full or partial duplication of the rhythmic cadences of the 1 ^st voice on the ^3rd voice. Such an implementation allows these two voices, throughout the piece or in certain passages, to play together but with different notes each. Otherwise, the rhythm clock generation operation of the voice 1 ^st (§ F1) applies identically to the ^3rd voice

A third operation consists in generating rhythmic cadences (mother cadences) for the fourth voice and possibly the third voice, according to the previous operation. The cadences produced here are based on variable and random lengths (from one time to several measures) and are thus different from the rhythmic cadences of each of the other voices. They are reproduced or not a random number of times at random or non-random locations but blank on the whole piece (cadence (s) girl (s)).

A density coefficient specific to each of the three additional voices is applied to the available note figures and is valued according to the desired style of music, the voice concerned and the current moment (1 ^st to 3 ^rd movement). For example, if the 1 ^st voice in the first movement has a density ratio of black and white greater than zero (all other figures of notes having a density coefficient equal to zero), the second voice of the first movement may have an eighth and / or sixteenth note density coefficient greater than zero.

A determination of the aeration of the rhythmic cadence then operates already described in § F1 (rhythmic rate of one voice ^AD).

Then, we proceed to a determination (random or not) of lengths) of silence of the locations to play of the rhythmic cadence of the ends of moments. This operation concerns, for each of the three movements, the rhythmic cadences of voices 1, 2, 3 and 4 already calculated. It allows a more or less long total “breathing” at the end of each moment (the whole orchestra stops).

G2 / Durations separating note attacks from voices 2, 3 and 4

The procedure is as for one voice ^age (see § F2). G3 / Family of playing positions for voices 2, 3 and 4

The family of notes that each position to be played must be determined: - according to the family of the possible note of the previous voice (s) already composed and located at the same position in the song; the notes of each voice in this position must belong to the same family, this condition having priority over all the following;

- according to its position in the current musical phrase; - according to its past and future neighborhood, that is to say the duration values which separate it from the previous playing position and the following playing position;

- possibly according to its rhythmic location in the current measure; - and according to the family of the previous note (see in particular the rule of alternation and its exceptions).

G4 / Degrees of pattern and patterns of voices 2, 3 and 4

Performing a similar operation to that described in § F5 to one voice ^era.

G5 / Pitch of voices 2, 3 and 4 (General)

It performs an operation similar to the one voice ^age (§ F6).

G6 / Intervals between the notes of voices 2, 3 and 4

It performs an operation similar to the one voice ^age (§ F7). G7 / Velocities of the notes of voices 2, 3 and 4

An operation similar to that for the 1 ^st voice (§ F9) is carried out.

G8 / game lengths notes voices 2, 3 and 4 is carried out an operation similar to that for the ^1st vote

(§ F10).

H / All Voices (1, 2, 3 and 4)

H1 / Shades of velocities of all voices

We proceed to a semi-random generation of nuances applicable to the velocities obtained by the previous operations, described in § F9 and § G7. All four voices are processed simultaneously. Certain passages of determined lengths are respectively of low, medium or high volume.

H2 / Crescendos and decrescendos of nuanced velocities of all voices

An operation of semi-random generation of crescendos and decrescendos is applied to the nuanced velocities of the previous operations (§ F9, G7 and H1), at the passages of lengths determined in the previous operation (cf. § H1). All four voices are processed simultaneously. 1 / Configuration of the midi soft-synthesizer

We send the different commands below to the soft synthesizer.

11 / General configuration

This is the GM configuration (for the “General MIDI” standard), thanks to which the synthesizer waits for commands and play orders from another device with the MIDI interface (in a computer).

12 / Program number (instrument)

It is defined for each MIDI channel used. Here, channels 1, 2, 3 and 4 are used respectively for voices 1, 2, 3 and 4.

13 / General effect number (reverb, delay, etc.)

14 / MIDI mix settings

These are the respective mix settings for the four MIDI channels (volume, pan, tune, etc.). These setting values for all composed voices are either arbitrarily defined or drawn at random from arbitrarily defined values.

J / Song play

In this step 100, the different values of the musical parameters of the game, calculated in the preceding operations, are sent to the soft-synthesizer.

The various operations to enhance the global musical parameters of an automatic counterpoint composition therefore include the steps (Figure 1):

- Initialization 101 of the pseudo-random draw function of the microprocessor;

- random draw 102 of the alphabetical title of the song to be composed;

- random choice 103 or not of the style of song to be composed;

- random draw 104 of its tone;

- draw 105 of the harmonic mode (minor or major); - random or non-random drawing 106 of harmonic sequences, that is to say chords of passages and their lengths;

- and draw 107 of the markers of harmonic modulation (s). The main stages in enhancing the musical parameters specific to each voice in automatic counterpoint composition include

(Figure 2) the prints 108, 109, 110 and 111 of the parameters of the music 1 ^st voice respectively, of the ^2nd voice, voice 3 ^rd and 4 ^th voice.

In addition, the steps of valuing the parameters of a single rhythmic cadence and of the corresponding note pitches, using the degree pattern process, include (Figure 3): - the random drawing 112 of a length of rhythmic cadence;

- the random drawing 113 of the positions to be played over the chosen length;

the determination 114 of the families of notes to be played at the positions of notes to be played; - the random drawing 115 of the degrees of pattern;

- the duplication 116 of the degrees of pattern;

- and the 117 valuation of the pitches of notes at the positions to be played according to the degrees of the father and son patterns, the general harmony and the current chord of passage.

LEXICON

"Accord" (musician) Or "chord of notes", or "harmony". Several different notes (or sounds), played simultaneously, represent a chord. Here, all the “notes” events of the melody and the accompanying instruments, except the percussion instruments, are calculated according to the current chord and the basic harmony (key).

"Agreement in progress" (internal definition)

The size of a piece is measured in time (universal). Each event, at a given time, is associated with other events. The current chord is one that has the same position in the song as a note in question.

Ex: On one of the positions of the melodic line of the song, we can read a note "d", and identify the current chord. "Passage agreement" (internal definition)

Passage chords are the different harmonic chords that succeed each other in a song. They can be played by one or more instruments (example rhythmic chords of a guitar) and they indicate a harmonic path traversed by a melody or to be traversed for the composition of one or more songs and of the orchestration. These chords are here those corresponding to the tones close to the general tone or general harmony defined by the armor or "key" (with their relative tones minor or major according to the initial harmonic mode), as well as the fundamental chord itself. This facilitates the harmonic modulation operation. The fundamental agreement is worth for example C or Am for an armor of C, which corresponds to nothing in the key (neither #, nor b). "Apprenticeship" (internal definition)

Applies to basic grades and passing grades, in certain application modes where alternation is imposed (with exceptions). "Anticipation" (internal definition)

Said of certain draws of positions to play or not, and of certain draws of note pitches. These draws are not made in chronological order in relation to the song positions. "Arpèges" (musician)

Part of the musical arrangement that can be interpreted by different instruments, depending on the style of music. An arpeggio represents a succession of monophonic notes from the family of base notes, such that two successive notes have between them a maximum number of intervals (semitones) chosen arbitrarily.

See also "Orchestra" and "Basic Notes". "Musical arrangement" (musician - arranger)

Original composition for each accompanying instrument according to a given melody. Here synonymous with "Orchestration". "Associate" (internal definition) Associated event, associated agreement.

Said of an event having the same position in the song. See "In progress". "Basic" (internal definition, musician) See “Note family”, “Basic notes”, “Basic harmonies”. "Cadence (s) girl (s)" (internal definition)

Copy (s) of a "Mother Cadence". See also "Son pattern". "Mother cadence" (internal definition)

Original random rhythmic cadence. See also "Father motif". "Rhythmic cadence" (musician)

Also called "Skeleton of the melody".

Designation of the positions which will receive a note (positions to be played). Help to understand the concept: "Hit with your hand the rhythmic rhythm of the song" In the moonlight "without singing it". See "Density".

The song is divided into positions (or time units). A random draw with constraints designates those positions which will receive notes (in fact the beginning of notes), and consequently those which will act as rests or durations of notes. "Singing" (musician)

Here always means "melody" or "melodic line". "Wide song" (musician)

Said of a melody composed mainly of rounds, white and black on a relatively slow tempo. "Consistent" (internal definition)

See "Consistent Music". "Constraints" (internal definition) See "Filters". "Counterpoint"

Art consisting of writing or playing several songs that are different from each other and yet pleasant to the ear. "Beginning of time" (internal definition)

The "location" of the beginning of time. That is to say "e1". See "Location". "Degrees of pattern" (internal definition) Not to be confused with the degrees of a diatonic scale or other harmonic scales. These are the degrees of a diatonic range amputated from the 6 ^th degree or the 7 ^th degree. Eg with a range of Cs: C, D, E, F, G, C, or: C, D, E, C, G, B, C.

Pattern degrees have values between 1 and n, the latter being the highest value accepted according to the range of the instrument called to play the voice being composed, and the value 1 corresponding to the lowest . The set of pattern degrees of a rhythmic cadence constitute a pattern. "Density" (internal definition)

Density coefficient.

It influences the possible number of notes per measure in the random drawing of rhythmic cadences, the musical flow being more or less dense or airy. A greater or lesser coefficient is applied to each location of the time of the measure of a piece according to the desired style of music. "Double-eighth note" (musician) Duration or length of note here equal to one unit of time.

"Doublons" (musician)

Successive repetition of the same note. See "Filters". "Note duration" (internal definition)

See “Note length” (musician). "Effects" (Sound engineer)

The “effects” are applied to the sounds (instruments): expression, vibrato, reverberation, glissando, pan, fine adjustment of the pitch of the sound (in comas). "Location" (internal definition)

"Position" of a note or silence within each beat.

In the mode of application described, there are four values of locations, called "e1", "e2", "e3", and "e4". “Passage location” and “Basic location” (internal definition) In the generation of rhythmic cadences, “passage locations” designates the locations located between the “basic” locations. The so-called “base” time slots are intended to receive only base notes, while the other so-called “passing” locations can receive passing notes or base notes or both, depending on the style to be obtained. In a rhythmic cadence, the basic locations alternate with the passage locations (with exceptions). Example with two locations per time: "e1" is a basic location, while "e2" is a passing location. "In progress" (internal definition)

Current events are those that have the same position in the song, so at a given time. Example: chord in progress during the "playing" of a note of the melody. See "Partner". "Enriched" (musician) Said of a passage agreement other than perfect. Examples: C9 #, Cm7 / 9. "State" (internal definition)

In a rhythmic cadence, defines a position. There are two possible states for a position:

- note start position: value = 1;

- position of silence (or note duration): value = 0. "E1" (internal definition)

First location of a time. "E2" (internal definition)

Second location of a time. "E3" (internal definition)

Third location of a time. "E4" (internal definition) Fourth and last location of a time.

"Note family"

See “Basic notes” and “Passing notes” “Filters” or “Constraints” (internal definition)

They intervene by imposing constraints on the random draws:

- rhythmic cadences of the melody;

- families of notes (base or passage) of the playing places defined by the rhythmic cadences;

- degrees of the motif of each mother cadence. ; - and the pitch of the melody notes.

Among the filters applied for the draws of note pitches, there are those indicated below. a / The interval filters (the intervals being expressed in numbers of semitones between two notes) reject the notes whose pitch is greater than a predefined interval compared to the previous note (except for the first) and compared to the next (except for the last). b / Duplicate regulation filters prohibit or not, depending on the desired style, the drawing of successive notes of the same pitch

(or duplicates). cl Harmonic filters include: d) state of the art chord filters; c2) filters for drawing the melody notes using the base notes and the passing notes; and c3) filters for drawing notes from the instruments accompanying the orchestras. "End of sentence" (internal definition)

Said of the last note of a musical phrase followed by at least three positions not to play (or in a variant, three positions of silence). "Truncated range" (internal definition)

It is about a range which does not respect the normal degrees of a diatonic range (degrees 1 to 7). It skips either the sixth degree or the seventh degree, the missing degree then being inserted or not in the melody line. This formatting of the degrees of the diatonic scale facilitates the rigorous alternation of the families of note in the drawing of the different pitches of note and this for all the voices of the song. "Germ" (computer scientist)

Germ of the pseudo-random draw of a microprocessor. This number can generally be between 1 and 2,147,483,647 and determines the successive random draws. "Group of locations" (internal definition) It is a group of “n” positions to play or not numbered from 1 to n. This group is reproduced throughout the song or to the next group of locations whose length (in number of positions) is different, and so on ... These rhythmic locations or positions are used to generate cadences rhythmic, defining the families of notes and drawing volumes for each note. The length of a group of pitches can be variable according to the desired style: a single beat, a measure, etc. It is however preferable, to obtain a rhythmically coherent music:

- that the length of each group is not less than the number of positions contained in a time,

- that, if multiple groups are generated, each group begins locations on the 1 ^st position of a measurement, - and that each group does not overlap two measurements, unless band length is a multiple of a measure . "Harmony" (musician) a- Art of assembling sounds (musician) b- Here, it is more precisely about the general harmony (tone) or else chords of passage. "Basic harmony" (musician, internal definition) or "Basic chord"

Used here to designate the tone in which we play: the tone, even if the reinforcement of the score remains unchanged (“#” or “b” in the key).

Example: C minor. "Melody note pitch" (musician)

The value, in MIDI language, can be between 57 and 77. "Clock" (computer scientist) Allows the timing of the sending of information to the MIDI sound module in binary word packets every nths of a second, according to the tempo drawn randomly for each song. “Instruments” (musician) or “Sounds” in MIDI language MIDI sound modules usually include a minimum of 128 different “sounds” numbered from 0 to 127. See “Orchestra”. "Interpretation" (internal definition)

During the "playing" phase, certain playing parameters are sent randomly to the instrument: expression, vibrato, glissando, etc. Thus, two repeated musical phrases each have a different color. This trick allows you to give a certain life to the game. "Introduction" or "First movement" (internal definition) This is a moment before the song. Its rhythmic and / or harmonic specifics may be different from those of other moments. "Game" (internal definition)

The program first carries out calculations, random draws of all the parameters constituting a musical work, then it sends these parameters in the form of MIDI codes (soft-synthesizer). It is this last phase which is called "play". See "Interpretation". In another embodiment, the game is executed as the composition in real time. "Large" (musician)

See "Wide song". "Line" (musician)

Suite of monophonic notes. "Bass line". "Melody line" (musician) See "Melody". "Note length" (musician)

Duration of the note. Example: the round lasts four beats, the eighth note half a beat. "Melody" (musician) or "Melody line" or "Song"

Refers here to a succession of monophonic notes calculated from a given chord structure. "Melodic" (internal definition)

Is said here, notes of the melody. Melodic notes. "Measure" (musician)

Dividing the length of the song.

Here, the measure contains four beats (4/4), or sixteen sixteenth notes maximum (one per unit of time). "Half time" (internal definition) The location halfway between the beginning and the end of time, ie "e3" (see "Location"). "MIDI" (musician - computer scientist)

The “MIDI” (Musical Instrument Digital Interface) or “Digital Musical Instrument Interface” system allows instruments with this interface to communicate with each other using communication cables (computers, synthesizers, and other digital musical instruments). Manufacturers of synthesizers and digital musical instruments have generally adopted this standard. In the embodiments described, the values of the different orders of note sets, as well as those of the parameters of instrumentation configurations and more generally of configuration of a polyphonic synthesizer, come from the General MIDI protocol. These parameters are sufficient to produce music with several instruments. There are thus, in MIDI, 16 polyphonic channels. For example, the Roland brand G800 synthesizer allows you to hear a maximum of 64 notes played simultaneously. However, the MIDI standard is only an intermediary between the music generator and the polyphonic instrument. In the case where a specific circuit is implemented, a circuit in which the music generator is part of the same electronic circuit as the polyphonic digital instrument, compliance with the MIDI standard is not essential.

"Moments of a song" (internal definition)

The rhythmic and / or harmonic specifics of a moment are generally different from those of other moments. The moments can consist of: ^1st movement, ^2nd movement, ^nth movement, introduction, verse, chorus, interlude, break, middle-part, bridge, chorus, final, coda, etc. "Piece" (internal definition)

Concept of time: entire song, entire musical work. "Motif" (internal definition) All the degrees of a range (here diatonic) applied to the different positions to play at the same rhythmic cadence. If it is a "mother" rhythmic cadence, its pattern is said to be "father" and if it is a "daughter" rhythmic cadence its pattern is said to be "son". "Coherent music" (internal definition)

In contrast to contemporary research music, where chance has far more rights than here, everything is governed by the laws of harmony and counterpoint and by invention itself. See “Random drawing with constraints”. "Note to play" (internal definition)

Unit of time or position or location of note to be played. The time units are randomly designated during the calculation of the rhythmic cadences, to receive pitch pitches. "Basic Notes" (internal definition)

Family of notes made up of the notes making up the current passage chord, duplicated from octave to octave according to the range of the instrument (the notion of "passage chord", that is to say, the chord played at a given time, should not be confused with that of "passing notes", these two concepts being completely distinct). The passage agreement can be perfect or enriched, that is to say "encrypted" or "composed". "Passing notes" (internal definition)

Family of notes made up of notes duplicated from octave to octave according to the range of the instrument, from the range of the general harmony in progress amputated by: the tonic, the third (minor or major as the case may be), and the fifth straight, decreased or increased as the case may be), of the associated passage agreement in progress, perfect or enriched (encrypted or composed). “Note-on” (MIDI)

Designates the start of a note. “Note-off” (MIDI) Indicates the end of the duration of a note.

"Orchestration" (musician-arranger)

See "Musical arrangement". "Orchestra" or "Style" (internal definition) The embodiments described involve several orchestras, the instruments (or possibly the instrument) playing the following parts:

- the first voice or melody or song; - the second voice, here with a low range and with a vocal response function;

- the third voice, of higher density than the two preceding voices with a tessitura equal to that of the first voice;

- and the fourth voice, higher than the previous three voices.

Each orchestra implements the four voices; it includes the mixing and effects of instruments or voices (if the orchestra uses only one instrument), as well as the value or limit values of the parameters characterizing a defined style. "Perfect" (musician - arranger)

Said of a passage chord comprising only the tonic, the third (minor or major) and the fifth (fair) of a diatonic scale.

"Parties" (musician - arranger) These are the different instruments or voices. See list of

"Parties" in "Orchestra", or the different voices of the counterpoint.

"Musical phrase" (musician)

Period of play between two periods of silence of defined length.

"Position" (internal definition)

Time scale unit of the song.

"First movement" (internal definition) See "Introduction". "Reversal" (musician - arranger)

Said of a chord whose notes are not in fundamental state, that is to say in ascending order from the fundamental note (example for the chord of Do (C): do, mi, sol) . The application mode here offers two different reversals and the fundamental state of each chord, which gives a certain realism to the rhythmic chord game (example: guitar). "Response to song" (musician)

Designates a melodic line (of the second voice for example) which plays when the main voice (song) does not play. This alternation of play gives an impression of musical dialogue between these two voices. The reverse of this principle is called "Tutti". "Resolution" (internal definition) See "Unit of time". "Semi-random" (internal definition)

Said draws of values of musical parameters subject to constraints or between two limit values: low and high. "Sounds" (musician) See "Instruments"

"Passage sub-agreement" (internal definition)

These are chords possibly inserted, depending on the desired style, within an already harmonized measure. These agreements are part of the ^3rd and / or 4 ^{th e} time to an extent four times, on the ^2nd and / or ^3rd time in a measure three times, and more generally on the start of the second half of the measurement for a so-called binary measurement. In the case of so-called "research" music, these chords can be located anywhere in the measure. "Structure of a song" (musician)

Number of measures, moments and repeats. "Style" (internal definition) See "Orchestra". "Time" (musician)

The time of the measurement, the measurement being divided into number of times.

In this case, four beats are used per measure. Thus, each beat contains four places for notes or rests, that is 4/4: the value of four sixteenth notes (one quarter note maximum).

See also "Location". "Tick" (internal definition)

Systematism translated here by arbitrary blockages of values of musical parameters.

“Random draw with constraints” (internal definition)

Drawing of numbers within a limited range (example: between 57 and 77), these numbers compared to others may or may not be accepted. We start again the drawing if the constraints require it. See "Filters".

"Neighbor Tones" (musician-arranger)

The summary below gives an idea of the concept of the word "Neighbor". For more information, we can refer to the rules of art. Neighboring tones are those whose key armor does not differ more than an alteration, as well as their minor relative tones. Example for the tone of C Major:

- relative minor tone of C: A minor; - neighboring tones: F Major (a flat key) and G Major (a sharp key);

- minor relative tones of the two chords above: D minor and E minor (with the same key armor respectively).

"Time unit" (internal definition) or "Resolution"

Equivalent here to a sixteenth note duration in note value, or a quarter of a sigh in silence value. A beat of a four-beat measure (4/4) contains four units, i.e. for a complete measure: sixteen time units (4 x

4 = 16). "Velocity" (musician - computer music)

Said, in MIDI language, the volume of the note. See "Volume". "Voices" See "Parties".

"Volume" (musician - music computer)

In MIDI language, applies to the general volume of the instrument and not to the notes themselves (see "Velocity").

Claims

1. Method of automatic musical generation, in which at least one succession of notes is produced in musical locations respectively, characterized in that game pointers are generated respectively in some of said musical locations, each of said game pointers indicating another said musical locations, so as to allow a non-sequential reading of said successions of notes.

2. A method of musical generation according to claim 1, characterized in that said successions of notes are generated in tones close to each other.

3. A method of musical generation according to one of claims 1 or 2, characterized in that said game pointers are placed at the start of measurements.

4. A method of musical generation according to one of claims 1 or 2, characterized in that a sequential reading of said successions of notes is provided after a reading by means of said playing pointers.

5. A method of musical generation according to any one of the preceding claims, characterized in that it comprises:

- at least one operation for defining (113 - 115) at least a succession of determined length of degrees of a range at determined locations to be played, called a pattern, - at least one duplication operation (116) of each of said patterns in free locations so as to form deduced patterns,

- And an operation for determining (117) the heights of notes of said patterns, as a function of the degrees of the patterns and of properties specific to each of the patterns.

6. A method of musical generation according to claim 5, characterized in that said specific properties include for each of said patterns a general harmony and a chord.

7. A method of musical generation according to any one of the preceding claims, characterized in that pieces consisting of sets of said successions of notes are generated, using prints of a combination of parameters for the generation of said pointers , and in that said method comprises an operation of pairing with each of said combinations of parameters, a unique digital identifier, said operation of pairing consisting in using (101) said identifier as seed for initialization of a pseudo-random printing apparatus .

8. An automatic musical generation device capable of producing at least one succession of notes in musical locations respectively, characterized in that said device comprises means for generating game pointers respectively in certain of said musical locations, each of said game pointers indicating another of said musical locations, so as to allow a non-sequential reading of said successions of notes, said device being preferably intended to implement the method according to any one of claims 1 to 7.

9. computer program product, characterized in that it comprises program code instructions for the execution of the operations of the music generation method according to any one of claims 1 to 7 when said program is executed on a computer .