Disclosure of Invention
In order to solve the above technical problems, it is an object of the present invention to provide a method, system, apparatus, and storage medium capable of automatically and rapidly generating a news video.
The first technical scheme adopted by the invention is as follows:
a news video generation method, comprising the steps of:
acquiring text news manuscript information;
analyzing the text news manuscript information to obtain text characteristics;
acquiring picture information and/or video information by combining the character features and a preset retrieval model;
and after generating voice information according to the text news manuscript information, generating a news video by combining the voice information and the picture information and/or the video information.
Further, the step of obtaining the character characteristics after analyzing the character newsletter information specifically includes the following steps:
identifying the title and the text of the news manuscript according to the character news manuscript information;
respectively identifying and acquiring noun words appearing in the title and the text, and calculating the appearance times of each noun word;
and combining preset weight standards and the occurrence times of each noun vocabulary to obtain a plurality of key noun vocabularies as character characteristics.
Further, the preset retrieval model is a web crawler model, and the step of acquiring the picture information and/or the video information by combining the character features and the preset retrieval model specifically comprises:
and scanning and retrieving in the network by combining the character characteristics and the web crawler model, and acquiring picture information and/or video information corresponding to the character characteristics.
Further, the step of generating the news video by combining the voice information and the picture information and/or the video information specifically includes the following steps:
typesetting the retrieved picture information and/or video information;
and synthesizing the voice information and the picture information and/or the video information into a news video by adopting a preset rendering engine.
Further, the step of synthesizing the voice information and the picture information and/or the video information into the news video by using a preset rendering engine specifically includes:
acquiring a playing scene model by combining character characteristics and a preset model database;
and synthesizing the voice information, the playing scene model and the picture information and/or the video information into a news video by adopting a preset rendering engine.
Further, the method also comprises a subtitle generating step, wherein the subtitle generating step specifically comprises the following steps:
and combining the text news manuscript information with a preset subtitle generator to generate subtitle information, and then fusing the subtitle information into a news video.
The second technical scheme adopted by the invention is as follows:
a news video generation method system comprises the following steps:
the data acquisition module is used for acquiring character news manuscript information;
the information analysis module is used for analyzing the character news manuscript information to obtain character characteristics;
the information retrieval module is used for acquiring picture information and/or video information by combining character characteristics and a preset retrieval model;
and the video synthesis module is used for generating a news video by combining the voice information and the picture information and/or the video information after generating the voice information according to the text news manuscript information.
Further, the information analysis module comprises an information splitting unit, a vocabulary counting unit and a characteristic obtaining unit;
the information splitting unit is used for identifying the title and the text of the news manuscript according to the character news manuscript information;
the word counting unit is used for respectively identifying and acquiring noun words appearing in the title and the text and calculating the occurrence frequency of each noun word;
the characteristic acquisition unit is used for acquiring a plurality of key noun vocabularies as character characteristics by combining preset weight standards and the occurrence frequency of each noun vocabulary.
The third technical scheme adopted by the invention is as follows:
a news video generation method and device comprises the following steps:
at least one processor;
at least one memory for storing at least one program;
when the at least one program is executed by the at least one processor, the at least one processor may implement a news video generation method as described above.
The fourth technical scheme adopted by the invention is as follows:
a storage medium having stored therein processor-executable instructions for performing the method as described above when executed by a processor.
The invention has the beneficial effects that: the method and the device automatically analyze and acquire the character characteristics according to the character news manuscript information, acquire the picture information and/or the video information according to the character characteristics, do not need to manually search and collect the picture or the video material, greatly save the time for collecting and editing the video material, improve the efficiency of making the news video and achieve the effect of quickly making the news video.
Detailed Description
As shown in fig. 1, the present embodiment provides a news video generation method, including the following steps:
s1, acquiring character newsletter information;
s2, analyzing the character newsletter information to obtain character characteristics;
s3, combining the character features and a preset retrieval model to obtain picture information and/or video information;
and S4, generating voice information according to the text news manuscript information, and then combining the voice information with the picture information and/or the video information to generate news video.
In the method of this embodiment, the text newsfeed information is a pure text newsfeed, and the text newsfeed information may be downloaded and acquired from the internet, for example, from various mainstream news websites such as the newseine, the civil network, the phoenix network, and the headline, and the text newsfeed information may be automatically acquired from the internet through program setting, or may be input after being searched and acquired by a user. After obtaining the text newsfeed information, the newsfeed is analyzed, and the news emphasis of the newsfeed is analyzed, so that corresponding text features are extracted, for example, by identifying keywords in a main title. According to the acquired character features, picture information and/or video information are acquired through a preset retrieval model, the preset retrieval model can be a web crawler model or a picture-text cross-modal retrieval model, the picture information is a picture corresponding to the character features, the video information is video data corresponding to the character features, for example, if the character features are bridges, pictures of a plurality of bridges or videos of aerial views of the bridges are acquired. And generating voice information according to the text news manuscript information, wherein the voice information is voice corresponding to the news manuscript, specifically, the voice information can be converted through a preset converter, and finally, combining the voice information with the picture information and/or the video information to generate a news video. Therefore, corresponding picture information and/or video information are/is automatically acquired according to the news manuscript, troubles such as video material collection and editing are avoided, the production time of the news video is greatly shortened, and the news video is rapidly generated.
Wherein, the step S2 specifically includes steps S21 to S23:
s21, identifying the title and text of the newsletter according to the character newsletter information;
s22, respectively identifying and acquiring noun words appearing in the title and the text, and calculating the appearance times of each noun word;
s23, combining the preset weight standard and the occurrence frequency of each noun vocabulary to obtain a plurality of key noun vocabularies as character characteristics.
In this embodiment, the title and the text of the newsfeed may be identified in various ways such as font and format, which may be implemented by using the existing technology and are not described herein again. After the title and the text are recognized, the noun words appearing in the title and the text are recognized respectively, for example, the noun words such as bridges, war and the like are recognized, the occurrence frequency of the words is counted respectively, and the keywords are obtained according to a preset weight standard. For example, a sports news-400 m free-play champion news is obtained about grand poplar, and the keywords are finally obtained by recognition: and if the grand poplar, 400 m, the free-tour game and the champion game are available, searching corresponding pictures or videos according to the obtained keywords, for example, obtaining pictures or videos of the grand poplar swimming, obtaining pictures or videos of the grand poplar boarding and receiving the prize, and the like.
The preset retrieval model is a web crawler model, and the step S3 specifically includes: and scanning and retrieving in the network by combining the character characteristics and the web crawler model, and acquiring picture information and/or video information corresponding to the character characteristics.
The preset retrieval model can be a web crawler model or a picture-text cross-modal retrieval model, when the picture-text cross-modal retrieval model is adopted, a picture-text database needs to be established in advance, and finally, the final picture information is obtained by comparing a similarity matrix of a picture and a text, the comparison result of the model is accurate, but a database needs to be established and the resource of the database is relatively limited, so the scheme adopts the web crawler model, the picture information and/or video information corresponding to the character characteristics is directly retrieved from the network through the web crawler model, the web crawler model is realized by adopting the existing model, and a special model structure is not required.
The step S4 specifically includes steps S41 to S43:
and S41, generating voice information according to the character news manuscript information.
And S42, typesetting the picture information and/or the video information obtained by searching.
And S43, synthesizing the voice information and the picture information and/or the video information into a news video by adopting a preset rendering engine.
In this embodiment, text-to-speech software may be used to generate speech information, or text newsletter information may be uploaded to a network platform for conversion, and then corresponding speech is downloaded. The picture information and/or the video information are typeset, wherein the typesetting can be performed manually by a user or automatically by a system. When the typesetting is performed manually, the playing sequence of each picture or video, the playing time of each picture or video and the like can be adjusted and set manually to generate continuous picture data. When the automatic typesetting is carried out, the system automatically sequences the pictures or the videos to generate continuous picture data. And finally, synthesizing the voice information and the picture information and/or the video information into a news video by adopting a rendering engine, wherein the generated news video can synchronously play the voice of the news manuscript on the audio and continuously play the corresponding pictures and/or videos on the video.
Referring to fig. 3 to 5, the step S43 specifically includes steps a1 to a 2:
a1, acquiring a playing scene model by combining character characteristics and a preset model database;
and A2, synthesizing the voice information, the playing scene model and the picture information and/or the video information into a news video by adopting a preset rendering engine.
In order to increase the richness of news playing, a playing scene model is set, the playing scene model is a virtual playing scene, for example, the playing scene model comprises a virtual host and a virtual playing background, the corresponding playing scene model is obtained according to character characteristics, for example, a news manuscript is a piece of news for a football match, after the character characteristics of the news manuscript are recognized as a football, the playing scene model corresponding to the football is obtained from a model database, for example, the host wears a football coat, a background picture is a football lawn, a virtual video playing window is arranged in the playing scene model and used for playing generated video data. Referring to fig. 3, the press release in fig. 3 is official news, and the corresponding play scene model is also serious, wherein the master is worn more officially. Referring to fig. 4, the newsfeed in fig. 4 is weather forecast news, and the virtual video playing window plays a corresponding weather forecast picture or video. Referring to fig. 5, the press release in fig. 5 is movie entertainment news, and the corresponding host has lively and casual dressing. The news video content is richer by setting various different playing scene models and providing different watching visions aiming at different types of news.
Further, as a preferred embodiment, the method further includes a subtitle generating step, where the subtitle generating step specifically includes:
and combining the text news manuscript information with a preset subtitle generator to generate subtitle information, and then fusing the subtitle information into a news video.
In this embodiment, the preset subtitle generator is used for generating subtitle information from the text newsfeed information, so that subtitles appear in the news video, and the watching experience of the news video is improved.
As shown in fig. 2, this embodiment further provides a news video generation method system, including:
the data acquisition module is used for acquiring character news manuscript information;
the information analysis module is used for analyzing the character news manuscript information to obtain character characteristics;
the information retrieval module is used for acquiring picture information and/or video information by combining character characteristics and a preset retrieval model;
and the video synthesis module is used for generating a news video by combining the voice information and the picture information and/or the video information after generating the voice information according to the text news manuscript information.
Further as a preferred embodiment, the information analysis module includes an information splitting unit, a vocabulary statistics unit and a feature acquisition unit;
the information splitting unit is used for identifying the title and the text of the news manuscript according to the character news manuscript information;
the word counting unit is used for respectively identifying and acquiring noun words appearing in the title and the text and calculating the occurrence frequency of each noun word;
the characteristic acquisition unit is used for acquiring a plurality of key noun vocabularies as character characteristics by combining preset weight standards and the occurrence frequency of each noun vocabulary.
The system for generating a news video, provided by the embodiment of the invention, can execute the method for generating a news video, can execute any combination of the implementation steps of the method embodiment, and has corresponding functions and beneficial effects of the method.
The embodiment also provides a news video generation method and device, including:
at least one processor;
at least one memory for storing at least one program;
when the at least one program is executed by the at least one processor, the at least one processor may implement a news video generation method as described above.
The news video generation method and device provided by the embodiment of the invention can execute the news video generation method provided by the embodiment of the method of the invention, can execute any combination of the implementation steps of the embodiment of the method, and have corresponding functions and beneficial effects of the method.
The present embodiments also provide a storage medium having stored therein processor-executable instructions, which when executed by a processor, are configured to perform the method as described above.
The storage medium provided by this embodiment may execute the news video generation method provided by the method embodiment of the present invention, may execute any combination of the implementation steps of the method embodiment, and has corresponding functions and advantages of the method.
While the preferred embodiments of the present invention have been illustrated and described, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.