Audio generation method and device, equipment and storage medium

By obtaining character attributes and scene information in literary works, and generating voice audio clips corresponding to multiple characters, the problem of difficult to realize the sound characteristics of multiple characters in the existing technology is solved, the listening experience and generation efficiency of audio literary works is improved, and the cost.

CN118942440APending Publication Date: 2024-11-12BEIJING YOUZHUJU NETWORK TECH CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202310538963.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-12
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The prior art is difficult to generate different sound characteristics for multiple characters in literary works, resulting in poor listening experience of audio novels, and high cost of manual dubbing, long periods and low efficiency.

Method used

By obtaining the character attributes of multiple characters in literary works, determining the target sound characteristics based on the character attributes, generating voice audio clips corresponding to the character, and combining multiple character audio clips into complete audio, using artificial intelligence technology to realize multiple roles Play.

Benefits of technology

It improves the listening sensory experience and comfort of audio literary works, realizes the auditory effect of multi-character show, reduces the production cost and cycle of audio generation, and improves the generation efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118942440A_ABST
    Figure CN118942440A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides an audio generation method and device, equipment and a storage medium. The method comprises the following steps: acquiring corresponding role attributes of a plurality of roles in the literature work and corresponding utterances taking the plurality of roles as main bodies; for each role in the plurality of roles, based on a role attribute of the role, determining a target sound characteristic for the role, and based on the target sound characteristic, generating a voice audio segment corresponding to a target utterance taking the role as a main body; and generating an audio for the literature at least based on the corresponding voice audio clips generated for the plurality of roles. Therefore, audios which contain different timbres and aim at the literature work can be conveniently and quickly generated based on different roles in the literature work, and the listening sensory experience of the generated audios is improved while the audio generation efficiency is ensured.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] Example embodiments of the present disclosure generally relate to the field of computers, and more particularly, to audio generation methods, devices, apparatuses, and computer-readable storage media. Background Art

[0002] With the rapid development of communication technology and smart devices, more and more applications provide users with various types of audio, which brings a lot of convenience to users. At present, some long audio products, such as audiobooks and blogs, have entered a stage of rapid development. For example, an audiobook application can play the audio corresponding to a book for users, so users can "read" the book without reading the book. In view of this, it is expected to generate corresponding audio for literary works such as books conveniently and quickly, so as to provide users with more audio products. Summary of the invention

[0003] In a first aspect of the present disclosure, an audio generation method is provided. The method comprises: obtaining corresponding role attributes of multiple roles in a literary work and corresponding utterances with the multiple roles as the main body; for each of the multiple roles, based on the role attributes of the role, determining a target sound characteristic for the role, and based on the target sound characteristic, generating a speech audio segment corresponding to a target utterance with the role as the main body; and generating audio for the literary work based on at least the corresponding speech audio segments generated for the multiple roles.

[0004] In a second aspect of the present disclosure, an audio generation device is provided. The device includes: an acquisition module configured to acquire corresponding role attributes of multiple roles in a literary work and corresponding speech with the multiple roles as the main body; a segment generation module configured to determine, for each of the multiple roles, a target sound characteristic for the role based on the role attribute of the role, and generate a speech audio segment corresponding to a target speech with the role as the main body based on the target sound characteristic; and a segment combination module configured to generate audio for the literary work based on at least the corresponding speech audio segments generated for the multiple roles.

[0005] In a third aspect of the present disclosure, an electronic device is provided. The device includes at least one processing unit; and at least one memory, the at least one memory is coupled to the at least one processing unit and stores instructions for execution by the at least one processing unit. When the instructions are executed by the at least one processing unit, the device executes the method of the first aspect.

[0006] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided, wherein a computer program is stored on the computer-readable storage medium, and the computer program can be executed by a processor to implement the method of the first aspect.

[0007] It should be understood that the contents described in this content section are not intended to limit the key features or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0009] Figure 1 A schematic diagram showing an example environment in which embodiments of the present disclosure can be implemented;

[0010] Figure 2 A schematic diagram showing an example architecture for audio generation according to some embodiments of the present disclosure;

[0011] Figure 3 A schematic diagram of an example of generating speech audio according to some embodiments of the present disclosure is shown;

[0012] Figure 4 A schematic diagram of an example of generating an additional audio segment according to some embodiments of the present disclosure is shown;

[0013] Figure 5 A flowchart showing a process of audio generation according to some embodiments of the present disclosure;

[0014] Figure 6 A schematic structural block diagram of an apparatus for audio generation according to some embodiments of the present disclosure is shown; and

[0015] Figure 7 A block diagram of an electronic device is shown in which one or more embodiments of the present disclosure may be implemented. DETAILED DESCRIPTION

[0016] It is understandable that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, scope of use, usage scenarios, etc. of the personal information involved in the present disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0017] For example, in response to receiving an active request from a user, a prompt message is sent to the user to clearly prompt the user that the operation requested to be performed will require obtaining and using the user's personal information. Thus, the user can autonomously choose whether to provide personal information to software or hardware such as an electronic device, application, server, or storage medium that performs the operation of the technical solution of the present disclosure according to the prompt message.

[0018] As an optional but non-limiting implementation, in response to receiving an active request from the user, the prompt information may be sent to the user in the form of a pop-up window, in which the prompt information may be presented in text form. In addition, the pop-up window may also carry a selection control for the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0019] It is understandable that the above notification and the process of obtaining user authorization are merely illustrative and do not constitute a limitation on the implementation of the present disclosure. Other methods that meet the relevant laws and regulations may also be applied to the implementation of the present disclosure.

[0020] It is understandable that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and relevant provisions.

[0021] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0022] It should be noted that the titles of any sections / subsections provided herein are not restrictive. Various embodiments are described throughout this article, and any type of embodiment may be included under any section / subsection. In addition, the embodiments described in any section / subsection may be combined in any manner with any other embodiments described in the same section / subsection and / or different sections / subsections.

[0023] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions may be included below. The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may be included below.

[0024] As used herein, the term "literary work" refers to a carrier that reflects a real or virtual society with words, pictures, images and / or other symbols. In the embodiments of the present disclosure, the genres of literary works include but are not limited to novels, scripts, fables, fairy tales, comics, etc.

[0025] As briefly mentioned above, it is expected that more literary works can be converted into corresponding audio to bring more convenience to users. Taking novels as an example, audio novels synthesized by artificial intelligence (AI) are developing in the direction of improving listening comfort and experience. Such audio novels are usually generated based on the content of the novel using Text To Speech (TTS) technology.

[0026] TTS technology can perform linguistic analysis on text content. Specifically, TTS technology can perform vocabulary, grammar and semantic analysis sentence by sentence to determine the underlying structure of the sentence and the composition of the phonemes of each word. Furthermore, the single words or phrases corresponding to the processed text can be extracted from the speech synthesis library, and the linguistic description can be converted into a speech waveform to generate the audio corresponding to the text.

[0027] Currently, some applications (for example, reading applications, audiobook applications, etc.) have provided audio novels that are broadcast in a single tone (also called unicast) using TTS technology. However, traditional solutions can only generate audio with a single tone, and cannot provide different characters in the novel with different styles of timbre. The audio novels generated in this way have a poor listening experience for users. On the other hand, the solution of manually dubbing literary works to generate audio can meet the requirements for the authenticity of the audio corresponding to literary works. For example, audio novels that can be broadcast by multiple characters (also called multicast) can be provided. However, the high production cost and long production cycle required for manual labor greatly limit the efficiency of audio generation.

[0028] The embodiments of the present disclosure propose a scheme for audio generation. According to various embodiments of the present disclosure, the role attributes of each of a plurality of characters and the speech of each character as the subject are obtained from a literary work. Based on the role attributes of each character, the voice characteristics (e.g., timbre) of the character are determined, and then the determined voice characteristics are used to generate a speech audio segment corresponding to the speech of the character. At least the speech audio segments corresponding to each of the plurality of characters are combined into audio for the literary work, also referred to as "audio literary work".

[0029] According to the embodiments of the present disclosure, in the production of audio literary works, different sound characteristics of characters with different attributes are taken into account. In the generated audio literary works, the auditory effect of multi-character performance is achieved. In addition, since the character attributes are taken into account, the listening sensory experience and comfort of the generated audio literary works are improved, and the accurate interpretation of the content of the literary works is achieved. In this way, originally silent literary works can be easily and quickly converted into audio products, thereby improving the generation efficiency of audio literary works. This helps to provide users with more and more vivid audio literary works.

[0030] Example Environment

[0031] Figure 1 1 is a schematic diagram of an example environment 100 in which embodiments of the present disclosure can be implemented. Figure 1 As shown, environment 100 may include an electronic device 110 that may acquire a literary work 120 for which audio is to be generated.

[0032] The literary work 120 depicts a plurality of characters, and each character has corresponding speech. It should be noted that the character can be any image with corresponding speech, including but not limited to people, animals, plants, objects, etc. For example, the character can be a talking potato or a talking dog, etc.

[0033] The literary work 120 may be provided to the electronic device 110 in any suitable format. In some embodiments, the content of the literary work 120 itself (e.g., the text of a novel, digital graphic data of a comic) may be provided to the electronic device 110. Alternatively or additionally, metadata describing the composition of the literary work 120 (if any) may be provided to the electronic device 110. Alternatively or additionally, information extracted from the literary work 120 (such as those described below) may be provided to the electronic device 110, and such information may be extracted from the literary work 110 by other devices or may be manually generated.

[0034] In the environment 100, the electronic device 110 may be any type of device with computing capabilities, including a terminal device or a server device. The terminal device may be any type of mobile terminal, fixed terminal or portable terminal, including a mobile phone, a desktop computer, a laptop computer, a notebook computer, a netbook computer, a tablet computer, a media computer, a multimedia tablet, a personal communication system (PCS) device, a personal navigation device, a personal digital assistant (PDA), an audio / video player, a digital camera / camcorder, a positioning device, a television receiver, a radio broadcast receiver, an electronic book device, a gaming device or any combination thereof, including accessories and peripherals of these devices or any combination thereof.

[0035] The server device can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks, and big data and artificial intelligence platforms. Server devices can include computing systems / servers such as mainframes, edge computing nodes, electronic devices in cloud environments, and so on.

[0036] The electronic device 110 may generate audio 130 for the literary work 120. In some embodiments, the electronic device 110 may utilize one or more trained machine learning models to implement audio generation. These trained models may be deployed in the electronic device 110, or deployed in other devices accessible to the electronic device 110.

[0037] The audio 130 is the audio corresponding to the literary work 120, also referred to as an audio literary work or an audio form of a literary work. The audio 130 includes at least audio corresponding to the speech of multiple characters in the literary work 120. In some embodiments, the audio 130 may also include audio corresponding to the narration in the literary work 120. Additionally, in some embodiments, the audio 130 may include an audio signal associated with the scene depicted by the literary work 120, such as ambient sound effects, background music, and the like.

[0038] It should be understood that the structure and functionality of the environment 100 are described for exemplary purposes only and do not imply any limitation on the scope of the present disclosure. The audio generation scheme according to the embodiments of the present disclosure may be implemented on any number of devices.

[0039] Some example embodiments of the present disclosure will be described below with continued reference to the accompanying drawings.

[0040] Example Audio Generation Architecture

[0041] Figure 2 A schematic diagram of an example architecture 200 for audio generation according to some embodiments of the present disclosure is shown. In general, the architecture 200 generally includes an information extraction unit 210, a sound characteristic assignment unit 220, a speech segment generation unit 230, and a segment combination unit 240. These units in the architecture 200 and other units not shown can be implemented in the electronic device 110.

[0042] The information extraction unit 210 is configured to obtain corresponding role attributes of multiple roles in the literary work 120 and corresponding discourses with multiple roles as subjects. Additionally, in some embodiments, the information extraction unit 210 can be further configured to obtain scene information 216 associated with scenes in the literary work 120. The scene information 216 describes the attributes of the scene and the events that occur in the scene. Specifically, the information extraction unit 210 can obtain role information 212, discourse information 214, and optional scene information 216 in the literary work 120, which are also collectively referred to as structured information of the literary work 120. That is, the information extraction unit 210 structures the literary work 120.

[0043] In some embodiments, the information extraction unit 210 may automatically extract structured information from the literary work 120. Natural language processing (NLP) technology and / or optical character recognition (OCR) technology, image recognition technology, etc. may be used to identify the content of the literary work 120 to extract structured information. For example, in the case where the literary work 120 is a pure text novel, the information extraction unit 210 may use any suitable NLP technology to determine structured information from the text of the literary work 120. For another example, in the case where the literary work 120 is a cartoon that combines text and graphics, the information extraction unit 210 may use OCR and NLP technology to extract structured information.

[0044] Alternatively or additionally, in some embodiments, the information extraction unit 210 may directly obtain the structured information from metadata describing the architecture of the literary work 120. For example, in the case where the literary work 120 is a script, the information extraction unit 210 may directly obtain the structured information from the script written in a predetermined format. Alternatively, the information extraction unit 210 may semi-automatically obtain the structured information. For example, the structured information may be determined from the manually annotated literary work 120 based on the manually annotated data.

[0045] Alternatively, in some embodiments, the architecture 200 may not include the information extraction unit 210. In such an embodiment, the electronic device 110 may directly obtain structured information generated manually, or obtain structured information provided by other devices.

[0046] The above describes an example of a method for obtaining structured information. The following describes structured information. The role information 212 includes a role identifier 201 and a role attribute 202 of each of the multiple roles in the literary work 120. The role identifier 212 may include, but is not limited to, the name of the role (e.g., Xiaoyi, Xiaoer, etc.), the identity of the role (e.g., Senior Brother, Master, etc.), the image of the role (e.g., different roles in comics have different images), and the like.

[0047] The character attributes 202 may include one or more attribute dimensions. Examples of attribute dimensions may include, but are not limited to, the age, gender, background, personality (e.g., character setting), camp (e.g., good guy, bad guy), etc. of the character in the literary work.

[0048] The role identification 201 and the role attribute 202 of each role in the role information 212 correspond to each other. Schematically, if the role attribute of role A needs to be obtained, the corresponding role attribute can be directly obtained from the role information 212 based on the identification of role A.

[0049] In some embodiments, different roles have different role identifiers 201, but different roles may have the same role attributes 202. For example, two roles are twins, and the role identifiers 201 of the two roles are different (for example, different names), but the role attributes 202 may be the same (for example, gender, age, personality, etc.).

[0050] The speech information 214 includes the narration 203 in the literary work 120 and the character speech 204 with multiple characters as the main body. The narration 203 is information other than the character speech 204, including but not limited to the description of the environment, the description of the plot, the description of the action, etc. The character speech 204 with a certain character as the main body may include but not limited to the dialogue between the character and other characters (for example, what the character says to another character), the inner monologue of the character, the soliloquy of the character, etc.

[0051] The above generally describes the role information 212 and the speech information 214. In order to more clearly understand the embodiments of the present disclosure, the following describes an example process of determining the role information 212 and the speech information 214. However, it should be understood that the described process is only exemplary and is not intended to limit the scope of the present disclosure.

[0052] The information extraction unit 210 can determine a set of candidate characters with names in the literary work 120 based on the text of the literary work 120. For example, any suitable named entity recognition (NER) technology can be used to determine the candidate characters. On the other hand, the information extraction unit 210 can determine the asides and character utterances in the literary work 120. For example, the novel can be split into narration and utterances using NLP technology. Further, the information extraction unit 210 can use a speaker determination model to determine the subject of each utterance.

[0053] In some embodiments, the information extraction unit 210 can use the subject of the identified speech to filter the extracted candidate roles. Specifically, for each candidate role in a group of candidate roles, the amount of speech with the candidate role as the subject can be determined. The candidate role whose amount of speech is greater than the threshold can be determined as one of the final multiple roles (that is, the role with the role identification 201). The amount of speech can be measured in any appropriate unit, such as the number of sentences, the number of words, etc. Exemplarily, since the literary work 120 may contain roles that only appear in other people's conversations and memories, such roles may have role attributes, but do not have speech with them as the subject. The information extraction unit 210 can filter out such roles without speech by setting a threshold value for the amount of speech, retain only the real roles with a certain amount of speech, and only extract the role information 212 corresponding to the role with a certain amount of speech. In this way, not only can the amount of data processing required for subsequent information extraction and allocation of sound characteristics be reduced, but also a more accurate interpretation of the content of the literary work can be achieved.

[0054] After determining the multiple roles in the literary work 120, the information extraction unit 210 can determine the role attributes 202 of each role. In some embodiments, the information extraction unit 210 can determine, for each of the multiple roles, the description information for the role and / or the relationship between the role and other roles in the multiple roles from the literary work. The description information can include, for example, the age, identity, and evaluation of the role by other roles in the literary work. The relationship can include, for example, the working relationship, family relationship, and friend relationship of the role in the literary work. Then, the role attributes of the role can be determined based on the description information and the relationship.

[0055] In some embodiments, the information extraction unit 210 may match the extracted description information and / or relationship with corresponding tags, and the tags are used to describe one or more dimensions of the character attributes. The information extraction unit 210 may use a trained tag matching model (e.g., a multi-classifier), or a pre-determined tag matching rule, or manual annotation information, etc., based on the description information and / or determine the tags associated with the character attributes corresponding to the character. These tags will be used as the character attributes 202 in the character information 212. For example, there may be a lot of descriptive information about the personality of a certain character in the literary work 120, and the information extraction unit 210 can match several corresponding tags based on these description information, thereby ensuring the accuracy and conciseness of the character attributes 202.

[0056] The above describes an example acquisition of role information 212 and speech signal 214. Continuing with architecture 200. Role information 212 is provided to a sound characteristic assignment unit 220. The sound characteristic assignment unit 220 is configured to determine, for each role in a plurality of roles, a target sound characteristic for the role based on the role attribute 202 of the role. In this context, the sound characteristic may include any suitable acoustic feature, such as timbre, pitch, loudness, or a combination thereof.

[0057] The speech information 214 is provided to the speech segment generation unit 230 together with the target sound characteristics corresponding to each of the multiple roles determined by the sound characteristic assignment unit 220. The speech segment generation unit 230 is configured to generate a speech audio segment corresponding to the target speech with the role as the subject based on the target sound characteristics. It can be understood that in the literary work 120, the speech with a certain role as the subject can include multiple segments, such as words in different scenes and words at different times in the same scene. In this case, the speech audio segment for the role can include sub-segments corresponding to the multiple segments respectively.

[0058] The corresponding speech audio segments generated for the multiple characters are provided to the segment combination unit 240. The segment combination unit 240 is configured to generate the audio 130 for the literary work 120 based at least on the corresponding speech audio segments generated for the multiple characters. That is, the segment combination unit 240 at least combines the speech audio segments into an audio literary work.

[0059] In the embodiment of the present disclosure, different roles are matched with voice characteristics (eg, timbre) that are adapted to their attributes, and automatic voice characteristic allocation is achieved. In this way, the effect of multi-person broadcasting can be achieved.

[0060] In some embodiments, in order to further improve the listening experience of the voice audio segment generated by the voice segment generation unit 230, the speech information 214 may also include the emotional characteristics 205 of the character's speech 204. For example, for the target speech of the character, the information extraction unit 210 may determine the emotional characteristics 205 contained in the target speech based on the content and context of the target speech. It should be noted that the emotions contained in different speeches may be different, and the emotions contained in speeches with the same content in different environments may be different. In addition, as mentioned above, a character's speech may include multiple paragraphs, and different emotions may be contained in each paragraph or even different sentences in a paragraph. Therefore, the information extraction unit 210 can determine the corresponding emotional characteristics 205 for each paragraph (or each sentence) of the character.

[0061] In this embodiment, the emotional characteristic 205 is also provided to the speech segment generation unit 230. The speech segment generation unit 230 accordingly converts the text of the target speech into a speech audio segment based on the target sound characteristic and the emotional characteristic 205. Compared with the speech audio segment generated based only on the target sound characteristic, the audio of the speech audio segment generated in combination with the emotional characteristic 205 can change with the change of emotion, so that the listening sensory experience of the audio literary work is higher.

[0062] Example of generating speech audio

[0063] Reference below Figure 3 An example embodiment is described of how speech audio segments are generated and combined into speech audio. Figure 3 An example architecture for generating speech audio sounds according to some embodiments of the present disclosure is shown. Figure 3 In the example, the matching unit 310 can be regarded as an example implementation of the sound characteristic assignment unit 220, the text-to-speech conversion unit 320 can be regarded as an example implementation of the speech segment generation unit, and the splicing unit 330 can be regarded as at least a part of the segment combination unit 240.

[0064] In this example, the role information 212 includes role attributes 301-1 of role 1, role attributes 301-2 of role 2, ..., role attributes 301-L of role L, etc., which are collectively or individually referred to as role attributes 301 of the role. L is a positive integer greater than or equal to 1.

[0065] The matching unit 310 obtains a plurality of candidate voice characteristics that are different from each other, each candidate voice characteristic corresponding to a role having a specific role attribute. Figure 3 As shown, the matching unit 310 can pre-acquire a sound characteristic library 315, and the sound characteristic library 315 includes multiple candidate sound characteristics, such as candidate sound characteristics 302-1, ..., 302-3, ..., 302-M, which are also collectively or individually referred to as candidate sound characteristics 302, where M is a positive integer greater than or equal to 1. In some embodiments, to ensure that the target sound characteristics determined for different roles are different, M is a positive integer greater than or equal to L. For each of the multiple roles, the matching unit 310 can select a target sound characteristic that matches the role attribute 301 of the role from the multiple candidate sound characteristics 302 based on the role attribute 301 of the role. Exemplary, as Figure 3 As shown, character 1 is matched to candidate voice characteristic 302 - 1 , character 2 is matched to candidate voice characteristic 302 - 3 , and character L is matched to candidate voice characteristic 302 -M .

[0066] In some embodiments, the matching unit 310 may use a trained matching model to determine the corresponding candidate voice characteristics 302 from the voice characteristics library 315 as the target voice characteristics based on the role attributes 301 of the role. The matching model may be a multi-classifier, which obtains classification results on multiple candidate voice characteristics 302 based on the role attributes 301 of the role. The candidate voice characteristics classified for a certain role are used as the target voice characteristics of the role. Figure 3 In the example in , after the matching unit 310 inputs the character attribute 301-1 of the character 1 into the matching model, the output result may be the candidate voice feature 302-1.

[0067] In some embodiments, the matching unit 310 can also determine the corresponding candidate sound characteristics 302 from the sound characteristic library 315 according to the character attributes 301 of the character based on a predetermined matching rule. The character attributes 301 of the character in the literary work may include attribute values ​​of multiple dimensions. The multiple dimensions may include but are not limited to the gender, age, camp, personality, etc. of the character in the literary work. The matching rule may indicate the corresponding relationship between the combination of attribute values ​​of multiple dimensions and the candidate sound characteristics 302. The matching rule may be manually pre-set or may be learned by the matching unit 310 based on a large amount of sound data. For each character, the matching unit 310 may select the candidate sound feature 302 corresponding to the combination of attribute values ​​of multiple dimensions of the character from the sound characteristic library 315 according to the matching rule as the target sound characteristic.

[0068] As an example, the character attributes 301-1 include attribute values ​​of multiple dimensions such as the character's age, personality, and camp. For example, the gender of character 1 is male, the age is young, the personality is high-cold and overbearing, and the camp is decent. The matching unit 310 can determine the target sound characteristics corresponding to character 1 as the sound characteristics with the label of "domineering president" based on the combination of these attribute values ​​based on the matching rules. For another example, the gender of character 2 is female, the age is old, the camp is decent, and the personality is kind and steady. The matching unit 310 can determine the target sound characteristics corresponding to character 2 as the sound characteristics with the label of "kind grandma" based on the combination of these attribute values ​​based on the matching rules. It should be understood that these attributes and sound characteristics are merely exemplary and are not intended to limit the scope of the present disclosure.

[0069] In this embodiment, an attribute system is established to label the characters in literary works, including information such as gender, age, camp, and personality. These tags can be combined to accurately describe character information. The voice feature library is a dubbing library composed of AI timbres of different styles and voices. According to the role-voice feature mapping rules, the best voice features, such as the best timbre, can be matched for characters with different attributes.

[0070] For each character, the text conversion unit 320 can convert the text of the target speech with the character as the main body into a voice audio segment corresponding to the target speech of the character based on the target sound characteristics of the character. Figure 3 In the example, the voice audio segment 303-1 of role 1, the voice audio segment 303-2 of role 2, ..., the voice audio segment 303-L of role L are generated, which are also collectively referred to as voice audio segments 303. The text-to-speech conversion unit 320 can use any suitable TSS technology to perform the conversion. In particular, as described above, in some embodiments, the speech information may include emotional characteristics. In such an embodiment, the text-to-speech conversion unit 320 can use emotional TTS technology to synthesize the voice audio segment.

[0071] The splicing unit 330 generates speech audio 304 from at least the speech audio segments 303 of these characters. In some embodiments, the text conversion unit 320 also generates speech audio segments corresponding to the narration. Accordingly, the splicing unit 330 also synthesizes the speech audio 304 with the speech audio segments 303 of the multiple characters. The splicing unit 330 can combine or splice according to the order in which the corresponding narration and speech appear in the literary work 120.

[0072] It is understandable that the speech audio segment 303 and the speech audio 304 only include synthesized human voices and can be regarded as dry voices. Specifically, the speech audio 304 can be regarded as a complete dry voice corresponding to the literary work 120.

[0073] In some embodiments, the speech audio 304 can be used as the audio 130 that the literary work 120 finally corresponds to. In other embodiments, the speech audio 304 is only a part of the audio 130, and the audio 130 also includes other audio signals.

[0074] Example of generating additional audio clips

[0075] Continue to refer Figure 2. As mentioned above, in some embodiments, the information extraction unit 210 can also obtain scene information 216 associated with the scene in the literary work 120. The scene information 216 describes the attributes of the scene and the events that occur in the scene. In some embodiments, the information extraction unit 210 can determine the scene and the content describing the scene in the literary work based on at least one of the time-space transformation or the role transformation in the literary work 120, and extract the scene information from the content. For example, the information extraction unit 210 can determine that the scene corresponding to the location A ends and the scene corresponding to the location B begins based on the text of the type "the character goes from location A to location B" that appears in the literary work 120. The information extraction unit 210 can use a trained scene division model to identify the scene information 216 from the literary work 120. Several literary works can be systematically annotated, and the annotated data can be used to train the scene division model. In this way, a scene and / or plot recognition system can be formed.

[0076] The scene information 216 includes at least one of the scene division 206, the plot 207, and the action 208. The scene division 206 is used to indicate the starting position and the ending position of the scene in the literary work 120, the environment and location of the scene, the time corresponding to the scene, etc. The plot 207 is the plot that occurs in the scene, such as fighting, quarreling, eating, meeting, etc. Action 208 refers to the action performed by the character in the scene. The scene information 216 is provided to the additional segment generation unit 250. The additional segment generation unit 250 is configured to generate an additional audio segment corresponding to the scene based on the scene information 216. It can be understood that the various sounds in the additional audio segment are matched or matched with the scene.

[0077] In this embodiment, the additional audio segment is provided to the segment combination unit 240. The segment combination unit 240 can generate the audio 130 for the literary work 120 based on the additional audio segment and the voice audio segment generated by the voice segment generation unit 230. That is, the segment combination unit 240 can combine the additional audio segment and the voice audio segment into an audio literary work. During the combination process, the segment combination unit 240 can regularize the volume of each audio segment so that the various sounds in the audio literary work are more reasonable, for example, the background music cannot dominate the main sound.

[0078] The additional segment generation unit 250 may store an additional audio library containing a plurality of additional sound audios. In some embodiments, the additional segment generation unit 250 may obtain a plurality of additional audio segments corresponding to a plurality of scenes based on the scene information 216 and the additional audio library using a trained machine learning model. In other embodiments, the additional audio segments may include sound effects and / or background music determined based on different types of scene attributes.

[0079] Reference below Figure 4 Such an example embodiment is described. Figure 4 An example architecture for generating additional audio segments according to some embodiments of the present disclosure is shown. Figure 4 In the example of FIG. 4 , the environment sound matching unit 410, the background music matching unit 420, and the action sound matching unit 430 can be regarded as Figure 2 An example implementation of the additional fragment generation unit 250 in FIG.

[0080] like Figure 4 As shown, the information extraction unit 210 can extract multiple scenes from the literary work 120, such as scenes 401-1, 401-2, ..., 401-N, which are also collectively or individually referred to as scenes 401, where N is a positive integer greater than or equal to 1. The scene information of each scene 401 can indicate at least one of the time of the scene, the location of the scene, the plot occurring in the scene, or the actions performed by the characters in the scene.

[0081] The ambient sound matching unit 410 is configured to determine the ambient sound corresponding to the time and location of the scene 401. For example, if the time of a scene is night and the location is the seaside, the ambient sound matching unit 410 can determine a sound effect with the sound of sea waves as the ambient sound of the scene.

[0082] The background music matching unit 420 is configured to determine the music corresponding to the plot occurring in the scene 401. For example, if the plot is that different characters are quarreling, the background music matching unit 420 may select urgent and tense music as the background music for this plot. For another example, if the plot is that a family is having a meal happily, the background music matching unit 420 may select warm music as the background music. It should be understood that different plots may occur in a scene, and appropriate background music may be selected for each plot.

[0083] The action sound effect matching unit 430 is configured to determine the action sound effect corresponding to the action performed by the character. For example, if the action is running, the action sound effect matching unit 430 can determine that the corresponding action sound effect is the sound of running. For another example, if the action is opening and closing a door, the action sound effect matching unit 430 can determine that the corresponding action sound effect is the sound of opening and closing a door. It should be understood that different actions may occur in a scene, and an appropriate action effect can be selected for each action.

[0084] As an example, the information extraction unit 210 can determine tags corresponding to the described scenes based on the text in the literary work, and use these tags as scene information. In this case, the environmental sound effect matching unit 410 can determine the matching environmental sound effects based on the tags corresponding to the combination of time and place, using predetermined matching rules or trained models. Similarly, the background music matching unit 420 determines the matching background music based on the tags corresponding to the plot, using predetermined matching rules or trained models. Similarly, the action sound effect matching unit 430 can determine the matching action sound effects based on the tags corresponding to the actions, using predetermined matching rules or trained models.

[0085] Furthermore, the environmental sound effects, background music, and action sound effects obtained by the environmental sound effect matching unit 410, the background music matching unit 420, and the action sound effect matching unit 430 are combined together as an additional audio segment corresponding to the scene 401. Figure 4 In the example, additional audio segment 402-1 of scene 401-1, additional audio segment 402-2 of scene 401-2, ..., additional audio segment 402-N of scene 401-N can be obtained, which are also collectively or individually referred to as additional audio segments 402.

[0086] continue Figure 4 The speech audio 305 is combined with the plurality of additional audio segments 402 to obtain the audio 130 corresponding to the literary work 120. This can be done, for example, by Figure 2 The segment combination unit 240 shown is implemented. The combination can be based on the order in which the corresponding words, plots, actions, etc. appear in the literary work.

[0087] In this embodiment, automatic post-production music / sound effects can be added, and background music, dynamic effects, ambient sounds, etc. can be matched according to the advancement of the storyline to achieve an effect that is not inferior to real-life audio. This can enhance the user's sense of immersion when listening to audio literary works.

[0088] Some example embodiments of the generation of audio literature works disclosed in the present invention are described above. In general, personalized voice audio for different characters can be achieved. In some embodiments, different sound effects are further matched for different scenes. In this way, high-quality audio can be obtained, which is conducive to improving the user's listening experience.

[0089] On the other hand, the embodiments of the present disclosure can solve the problems of high production cost and long production cycle of live multicast audio. For example, a multicast audio novel of a million-word novel can be produced with only very low TTS synthesis cost and a few minutes of synthesis time. Therefore, the use of machine learning and other technologies can achieve high-efficiency and low-cost mass production of high-quality audio literature works.

[0090] Example Process

[0091] Figure 5 FIG. 5 is a flowchart of a process 500 for audio generation according to some embodiments of the present disclosure. The process 500 may be implemented at the electronic device 110. Figure 1 Process 500 is described.

[0092] In block 510 , the electronic device 110 obtains corresponding role attributes of a plurality of roles in a literary work and corresponding utterances with the plurality of roles as subjects.

[0093] In box 520, the electronic device 110 determines, for each of the multiple roles, a target voice characteristic for the role based on the role attributes of the role, and generates a speech audio segment corresponding to a target utterance with the role as the subject based on the target voice characteristic.

[0094] At block 530 , the electronic device 110 generates audio for the literary work based at least on the corresponding voice audio segments generated for the plurality of characters.

[0095] In some embodiments, determining the target voice characteristic includes: obtaining a plurality of candidate voice characteristics that are different from each other, each candidate voice characteristic corresponding to a character having a specific character attribute; and selecting the target voice characteristic from the plurality of candidate voice characteristics based on the character attribute of the character.

[0096] In some embodiments, the role attribute of the role includes attribute values ​​of multiple dimensions, and selecting the target voice characteristic includes: selecting a candidate voice feature corresponding to a combination of the attribute values ​​of the multiple dimensions from multiple candidate voice characteristics as the target voice characteristic.

[0097] In some embodiments, generating the speech audio segment includes: determining the emotional characteristics contained in the target speech based on the content and context of the target speech; and converting the text of the target speech into the speech audio segment based on the target sound characteristics and emotional characteristics.

[0098] In some embodiments, process 500 also includes: obtaining scene information associated with a scene in a literary work, the scene information describing attributes of the scene and events occurring in the scene; and based on the scene information, generating an additional audio clip corresponding to the scene, and wherein generating audio for the literary work is further based on the additional audio clip.

[0099] In some embodiments, the scene information indicates at least one of the following: the time of the scene, the location of the scene, the plot occurring in the scene, or the actions performed by the characters in the scene.

[0100] In some embodiments, generating the additional audio segment includes combining two or more of the following into the additional audio segment: environmental sound effects corresponding to time and place, music corresponding to the plot, or action sound effects corresponding to the action.

[0101] In some embodiments, acquiring scene information includes: determining the scene and content describing the scene in the literary work based on at least one of a time-space transformation or a role transformation in the literary work; and extracting the scene information from the content.

[0102] In some embodiments, multiple roles are determined as follows: based on the text of the literary work, determining a set of candidate roles with names in the literary work; for each candidate role in the set of candidate roles, determining the amount of discourse in the literary work with the candidate role as the subject; and determining the candidate role whose amount is greater than a threshold as one of the multiple roles.

[0103] In some embodiments, obtaining corresponding role attributes of multiple roles includes: for a given role among the multiple roles, determining, from a literary work, descriptive information for the given role and the relationship between the given role and other roles among the multiple roles; and determining the role attributes of the given role based on the descriptive information and the relationship.

[0104] Example devices and equipment

[0105] Figure 6 A schematic structural block diagram of an apparatus 600 for audio generation according to some embodiments of the present disclosure is shown. The apparatus 600 may be implemented as or included in the electronic device 110. Each module / component in the apparatus 600 may be implemented by hardware, software, firmware or any combination thereof.

[0106] As shown in the figure, the device 600 includes an acquisition module 610, which is configured to acquire corresponding role attributes of multiple roles in a literary work and corresponding speech with the multiple roles as the main body. The device 600 also includes a segment generation module 620, which is configured to determine the target sound characteristics for each of the multiple roles based on the role attributes of the role, and generate a speech audio segment corresponding to the target speech with the role as the main body based on the target sound characteristics. The device 600 also includes a segment combination module 630, which is configured to generate audio for the literary work based on at least the corresponding speech audio segments generated for the multiple roles.

[0107] In some embodiments, the segment generation module 620 is further configured to: obtain a plurality of candidate sound characteristics that are different from each other, each candidate sound characteristic corresponding to a character having specific character attributes; and select a target sound characteristic from the plurality of candidate sound characteristics based on the character attributes of the character.

[0108] In some embodiments, the role attributes of the character include attribute values ​​of multiple dimensions, and the segment generation module 620 is further configured to select a candidate sound feature corresponding to a combination of attribute values ​​of multiple dimensions from multiple candidate sound characteristics as the target sound characteristic.

[0109] In some embodiments, the segment generation module 620 is further configured to: determine the emotional characteristics contained in the target speech based on the content and context of the target speech; and convert the text of the target speech into a speech audio segment based on the target sound characteristics and emotional characteristics.

[0110] In some embodiments, the device 500 also includes: a scene information acquisition module, configured to acquire scene information associated with a scene in a literary work, the scene information describing the attributes of the scene and events occurring in the scene; and an additional audio generation module, configured to generate an additional audio segment corresponding to the scene based on the scene information, and wherein the generation of audio for the literary work is further based on the additional audio segment.

[0111] In some embodiments, the scene information indicates at least one of the following: the time of the scene, the location of the scene, the plot occurring in the scene, or the actions performed by the characters in the scene.

[0112] In some embodiments, generating the additional audio segment includes combining two or more of the following into the additional audio segment: environmental sound effects corresponding to time and place, music corresponding to the plot, or action sound effects corresponding to the action.

[0113] In some embodiments, the scene information acquisition module is further configured to: determine the scene and the content describing the scene in the literary work based on at least one of the time-space transformation or the role transformation in the literary work; and extract the scene information from the content.

[0114] In some embodiments, multiple roles are determined as follows: based on the text of the literary work, determining a set of candidate roles with names in the literary work; for each candidate role in the set of candidate roles, determining the amount of discourse in the literary work with the candidate role as the subject; and determining the candidate role whose amount is greater than a threshold as one of the multiple roles.

[0115] In some embodiments, the acquisition module 610 is further configured to: determine, for a given role among multiple roles, descriptive information for the given role and the relationship between the given role and other roles among the multiple roles from a literary work; and determine role attributes of the given role based on the descriptive information and the relationship.

[0116] Figure 71 shows a block diagram of an electronic device 700 in which one or more embodiments of the present disclosure may be implemented. It should be understood that Figure 7 The electronic device 700 shown is merely exemplary and should not constitute any limitation on the functionality and scope of the embodiments described herein. Figure 7 The electronic device 700 shown can be used to implement Figure 1 An electronic device 110.

[0117] like Figure 7 As shown, the electronic device 700 is in the form of a general electronic device. The components of the electronic device 700 may include, but are not limited to, one or more processors or processing units 710, a memory 720, a storage device 730, one or more communication units 740, one or more input devices 750, and one or more output devices 760. The processing unit 710 may be an actual or virtual processor and is capable of performing various processes according to a program stored in the memory 720. In a multi-processor system, multiple processing units execute computer executable instructions in parallel to improve the parallel processing capability of the electronic device 700.

[0118] The electronic device 700 typically includes a plurality of computer storage media. Such media may be any accessible media that is accessible to the electronic device 700, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 720 may be a volatile memory (e.g., a register, a cache, a random access memory (RAM)), a non-volatile memory (e.g., a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. The storage device 730 may be a removable or non-removable medium, and may include a machine-readable medium, such as a flash drive, a disk, or any other medium, which may be capable of being used to store information and / or data (e.g., training data for training) and may be accessed within the electronic device 700.

[0119] The electronic device 700 may further include additional removable / non-removable, volatile / non-volatile storage media. Figure 7 As shown in , a disk drive for reading or writing from a removable, non-volatile disk (e.g., a "floppy disk") and an optical drive for reading or writing from a removable, non-volatile optical disk may be provided. In these cases, each drive may be connected to the bus (not shown) by one or more data media interfaces. The memory 720 may include a computer program product 725 having one or more program modules that are configured to perform various methods or actions of various embodiments of the present disclosure.

[0120] The communication unit 740 implements communication with other electronic devices through a communication medium. Additionally, the functions of the components of the electronic device 700 can be implemented with a single computing cluster or multiple computing machines that can communicate through a communication connection. Therefore, the electronic device 700 can operate in a networked environment using a logical connection with one or more other servers, a network personal computer (PC), or another network node.

[0121] The input device 750 may be one or more input devices, such as a mouse, a keyboard, a tracking ball, etc. The output device 760 may be one or more output devices, such as a display, a speaker, a printer, etc. The electronic device 700 may also communicate with one or more external devices (not shown) through the communication unit 740 as needed, such as a storage device, a display device, etc., communicate with one or more devices that allow a user to interact with the electronic device 700, or communicate with any device that allows the electronic device 700 to communicate with one or more other electronic devices (e.g., a network card, a modem, etc.). Such communication may be performed via an input / output (I / O) interface (not shown).

[0122] According to an exemplary implementation of the present disclosure, a computer-readable storage medium is provided, on which computer-executable instructions are stored, wherein the computer-executable instructions are executed by a processor to implement the method described above. According to an exemplary implementation of the present disclosure, a computer program product is also provided, which is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions, and the computer-executable instructions are executed by a processor to implement the method described above.

[0123] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices, equipment, and computer program products implemented according to the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0124] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0125] Computer-readable program instructions can be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, so that the instructions executed on the computer, other programmable data processing apparatus, or other device implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0126] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple implementations of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of a module, program segment or instruction includes one or more executable instructions for realizing the logical function of the specification. In some implementations as replacements, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be realized by a special hardware-based system that performs the function or action of the specification, or can be realized by a combination of special hardware and computer instructions.

[0127] The above descriptions of various implementations of the present disclosure are exemplary, non-exhaustive, and not limited to the disclosed implementations. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The selection of terms used herein is intended to best explain the principles of the implementations, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the various implementations disclosed herein.

Claims

1. An audio generation method, comprising: Obtaining corresponding role attributes of a plurality of roles in a literary work and corresponding discourses with the plurality of roles as subjects; For each of the plurality of roles, Based on the character attributes of the character, a target sound characteristic for the character is determined, and Based on the target sound characteristics, generating a speech audio segment corresponding to a target speech with the character as the main subject; as well as Audio for the literary work is generated based at least on the corresponding speech audio segments generated for the multiple characters.

2. The method of claim 1 , wherein determining the target sound characteristic comprises: obtaining a plurality of candidate voice characteristics that are different from each other, each candidate voice characteristic corresponding to a character having a specific character attribute; as well as The target voice characteristic is selected from the plurality of candidate voice characteristics based on the character attribute of the character.

3. The method according to claim 2, wherein the character attributes of the character include attribute values ​​of multiple dimensions, and selecting the target voice characteristics comprises: A candidate sound feature corresponding to a combination of the attribute values ​​of the plurality of dimensions is selected from the plurality of candidate sound features as the target sound feature.

4. The method of claim 1, wherein generating the speech audio segment comprises: Determining the emotional characteristics contained in the target utterance based on the content and context of the target utterance; as well as Based on the target sound characteristics and the emotional characteristics, the text of the target utterance is converted into the speech audio segment.

5. The method according to claim 1, further comprising: Acquire scene information associated with a scene in the literary work, the scene information describing attributes of the scene and events occurring in the scene; as well as Based on the scene information, an additional audio segment corresponding to the scene is generated, and Wherein generating the audio for the literary work is further based on the additional audio segment.

6. The method according to claim 5, wherein the scene information indicates at least one of the following: The time of the scene, The location of the scene, what happens in the scene, or Actions performed by characters in the scene.

7. The method of claim 6, wherein generating the additional audio segment comprises combining two or more of the following into the additional audio segment: Ambient sound effects corresponding to said time and said place, music that corresponds to the plot, or Action sound effects corresponding to the actions described.

8. The method according to claim 5, wherein acquiring the scene information comprises: Determining the scene and content describing the scene in the literary work based on at least one of a time-space transformation or a role transformation in the literary work; as well as The scene information is extracted from the content.

9. The method according to claim 1, wherein the plurality of roles are determined by: Based on the text of the literary work, determining a set of candidate characters with names in the literary work; For each candidate character in the set of candidate characters, determining the amount of discourse in the literary work with the candidate character as the subject; as well as The candidate role for which the amount is greater than a threshold is determined as one of the plurality of roles.

10. The method according to claim 1, wherein obtaining corresponding role attributes of the plurality of roles comprises: For a given role among the plurality of roles, Determining, from the literary work, descriptive information for the given character and a relationship between the given character and other characters in the plurality of characters; as well as Based on the description information and the relationship, role attributes of the given role are determined.

11. An audio generating device, comprising: An acquisition module is configured to acquire corresponding role attributes of a plurality of roles in a literary work and corresponding discourses with the plurality of roles as subjects; A fragment generation module is configured to generate a fragment for each of the plurality of roles. Based on the character attributes of the character, determine the target sound characteristics for the character, and Based on the target sound characteristics, generating a speech audio segment corresponding to a target speech with the character as the main subject; as well as The segment combination module is configured to generate audio for the literary work based on at least the corresponding voice audio segments generated for the multiple characters.

12. An electronic device comprising: at least one processing unit; as well as At least one memory, the at least one memory being coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions causing the electronic device to perform the method according to any one of claims 1 to 10 when executed by the at least one processing unit.

13. A computer-readable storage medium having a computer program stored thereon, wherein the computer program can be executed by a processor to implement the method according to any one of claims 1 to 10.

Citation Information

Cited By

  • System and method for automatically generating audiobooks with animation and computer-readable storage medium thereof

    TWI929632B

  • Method and apparatus for synthesizing audio, device, and product

    WO2026171218A1