A Method and System for Automatic Program Generation Based on Program Timeline and Multi-Agent Collaboration

Abstract

This invention relates to a method and system for automatic program generation based on a program timeline and multi-agent collaboration. The method includes: a content generation agent generating a program script based on program parameters; a sound generation agent automatically matching broadcast timbre according to the program script to generate human voice audio and determine background music and sound effects; a program synthesis agent splicing and synthesizing the human voice audio, background music, and sound effects according to a preset order and dependency relationship based on the program timeline to obtain program audio and a timeline file; and a review agent reviewing the program audio, triggering the corresponding agent to perform targeted backtracking and partial regeneration of the corresponding type of smallest content unit based on review tags, and performing partial replacement in the timeline file while keeping other smallest content units unchanged, thus completing the automated generation of broadcast audio programs. This method can achieve high-efficiency, high-consistency, and high-compliance program production while reducing production costs.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and system for automatic program generation based on program timeline and multi-agent collaboration. Background Technology

[0002] With the continuous expansion of radio and television, converged media platforms, and online audio application scenarios, the demand for radio audio program production continues to grow. Currently, radio audio program production generally adopts a segmented process that is mainly manual with the assistance of tools. This process mainly includes multiple stages such as topic selection and planning, data collection, scriptwriting, anchor recording, background music and sound effects selection, post-production mixing and editing, and manual content review and quality verification.

[0003] Traditional methods of producing radio audio programs have significant limitations: First, they are highly reliant on manual labor: script planning, style control, broadcasting method selection, and post-production rely heavily on human experience, resulting in long production cycles and high costs. Second, they have low automation: existing text generation or TTS technologies only cover single stages, lacking end-to-end automation solutions from "content generation—sound synthesis—program synthesis—review." Third, they lack personalization and style adaptation capabilities: it is difficult to automatically generate multiple versions of scripts and matching broadcast voices, background music, and sound effects based on program themes, durations, and styles. Fourth, they pose high quality and compliance risks: traditional processes rely heavily on manual spot checks for pronunciation accuracy, sensitive words, emotional nuances, and audio quality, which carries the risk of omissions and fails to meet broadcast-level requirements.

[0004] Therefore, traditional radio audio program production methods often suffer from high production costs and low efficiency because each stage relies heavily on human experience and lacks a unified data structure and collaborative mechanism between script planning, sound synthesis, audio production, and quality review. Summary of the Invention

[0005] In order to solve the above-mentioned technical problems, a method and system for automatic program generation based on program timeline and multi-agent collaboration is provided, which can achieve high-efficiency, high-consistency and high-compliance program production, and reduce production costs.

[0006] A method for automatic program generation based on program timeline and multi-agent collaboration, the method comprising:

[0007] The content generation agent obtains intent information, determines program parameters based on the intent information, and generates program scripts based on the program parameters.

[0008] The voice-generating intelligent agent automatically matches the broadcast timbre according to the program script, generates human voice audio, and determines background music and sound effects.

[0009] Based on the program timeline, the program synthesis agent splices and synthesizes the human voice audio, background music and sound effects according to a preset order and dependency relationship to obtain the program audio and timeline files.

[0010] The program audio is reviewed by an intelligent review agent, which generates review tags and maps them to the timeline file.

[0011] The review agent triggers the corresponding agent to perform targeted backtracking and partial regeneration of the smallest content unit of the corresponding type according to the review tag, and performs partial replacement in the timeline file while keeping other smallest content units unchanged, thus completing the automated generation of broadcast audio programs;

[0012] The smallest content unit includes the smallest unit of program script, the smallest unit of human voice audio, the smallest unit of background music, and the smallest unit of sound effects.

[0013] In one embodiment, an intent information is obtained through a content generation agent, program parameters are determined based on the intent information, and program script is generated based on the program parameters, including:

[0014] The content-generating intelligent agent performs semantic parsing and intent recognition through multi-round human-computer interactive dialogue to obtain intent information including program theme, type, audience, and scene;

[0015] The content generation agent configures program parameters, including target duration, copywriting style, broadcasting format, and emotional tone, based on the intent information.

[0016] The content generation agent calls a large language model to summarize materials for the program theme and form structured content materials;

[0017] The content generation agent generates program scripts based on the program parameters and the structured content materials, and then breaks down the program scripts into the smallest units of program scripts.

[0018] In one embodiment, a voice generation agent automatically matches the broadcast timbre according to the program script, generates human voice audio, and determines background music and sound effects, including:

[0019] The voice generation agent performs semantic analysis and emotion recognition on the program script to obtain the script style, and matches the broadcast format and corresponding virtual anchor voice according to the script style;

[0020] The voice generation agent determines the speech synthesis control parameters based on the emotion recognition results of the program script, calls the speech synthesis engine, and generates human voice audio corresponding to the emotion recognition results.

[0021] The sound generation agent selects or generates background music and sound effects that match the plot based on the semantic analysis results of the program script, and breaks down the background music and sound effects into each smallest unit of background music and the smallest unit of sound effects.

[0022] In one embodiment, a program synthesis agent, based on the program timeline, splices and synthesizes the human voice audio, background music, and sound effects according to a preset sequential dependency relationship to obtain a program audio and timeline file, including:

[0023] The program synthesis intelligent agent uses the program timeline as the core data structure. According to the sequential dependency relationship of program script, human voice audio, background music and sound effects, it uses a relative timestamp and anchor point alignment mechanism to align each of the smallest content units on the timeline.

[0024] The program synthesis agent calls the audio synthesis engine to mix the human voice audio, background music and sound effects to obtain program audio and structured timeline files.

[0025] In one embodiment, the program audio is reviewed by a review agent, review tags are generated and mapped to the timeline file, including:

[0026] The review agent performs multi-dimensional review and detection on the program audio and obtains the detection results; the multi-dimensional review and detection includes pronunciation accuracy detection, sensitive word detection, emotion consistency detection and audio quality detection;

[0027] The auditing agent generates structured auditing tags based on the detection results;

[0028] The review agent maps the review tags to the corresponding smallest content unit in the timeline file, adds review tag nodes to the timeline file, establishes a unique association between the review tags and the smallest content unit, and stores it in the form of structured data.

[0029] In one embodiment, the review agent performs multi-dimensional review and detection on the program audio to obtain the detection results, including:

[0030] The pronunciation accuracy detection uses a speech recognition engine to convert human voice audio into recognized text, compares the recognized text with the program script segment by segment, identifies pronunciation errors, omissions or repetitions, and maps the problem location to the corresponding smallest unit of human voice audio and the smallest unit of program script.

[0031] The sensitive word detection matches the program text and speech recognition results using a sensitive word database, and combines the contextual analysis of a large language model to identify sensitive word issues and map the problem location to the corresponding smallest unit of the program text.

[0032] The emotion consistency detection analyzes the actual emotion characteristics of human voice audio through a voice emotion recognition model and compares them with the preset emotion tags of each segment recorded in the timeline file. When the emotion matching degree is lower than the set threshold, it is determined to be an emotion mismatch problem, and the problem location is mapped to the corresponding smallest unit of human voice audio and the smallest unit of program script.

[0033] The audio quality detection performs volume detection, signal-to-noise ratio detection, distortion detection, and spectrum analysis on the program audio to identify problems such as abnormal volume, abnormal mixing, or abnormal sound quality, and maps the problem location to the smallest unit of background music and the smallest unit of sound effects in the corresponding time interval.

[0034] In one embodiment, the review agent triggers the corresponding agent to perform targeted backtracking and partial regeneration of the smallest content unit of the corresponding type based on the review tag, and performs partial replacement in the timeline file while keeping other smallest content units unchanged, thereby completing the automated generation of broadcast audio programs, including:

[0035] The review agent parses the review tag to obtain the problem type and the smallest content unit identifier, and determines the agent to be triggered and the smallest content unit to be traced back based on the problem type and the smallest content unit identifier.

[0036] Trigger the content generation agent, sound generation agent, or program synthesis agent to locally regenerate the smallest content unit of the corresponding type;

[0037] The program synthesis agent uses a relative timestamp and anchor point alignment mechanism to partially replace the smallest content unit corresponding to the review tag in the program timeline, while keeping the timing and content of the remaining smallest content units unchanged, thus completing the automated generation of broadcast audio programs.

[0038] In one embodiment, triggering a content generation agent, a sound generation agent, or a program synthesis agent to locally regenerate the smallest content unit of the corresponding type includes:

[0039] When the problem type is inaccurate pronunciation or omission, the review agent only triggers the sound generation agent, which obtains the text content corresponding to the review tag and regenerates the human voice audio.

[0040] When the issue type in the review tag is "emotion mismatch", the review agent triggers the content generation agent to rewrite the text content corresponding to the review tag, generate new text, and triggers the sound generation agent to regenerate human voice audio based on the rewritten text.

[0041] When the issue type in the review tag is a sensitive word, the review agent triggers the content generation agent to rewrite the text content corresponding to the review tag in a compliant manner, generates a new program script, and triggers the sound generation agent to regenerate human voice audio based on the rewritten script.

[0042] When the problem type in the review label is abnormal volume, abnormal mixing, or abnormal sound quality, the review agent triggers the program synthesis agent to determine the time interval corresponding to the text content of the review label, identify the smallest content unit within the time interval, recalculate the mixing parameters, and perform local remixing.

[0043] An automatic program generation system based on program timeline and multi-agent collaboration, the system comprising:

[0044] A content generation agent is used to acquire intent information, determine program parameters based on the intent information, and generate program scripts based on the program parameters.

[0045] A voice generation intelligent agent is used to automatically match the broadcast timbre according to the program script, generate human voice audio, and determine background music and sound effects;

[0046] The program synthesis intelligent agent is used to splice and synthesize the human voice audio, background music and sound effects according to a preset order dependency relationship based on the program timeline to obtain program audio and timeline files;

[0047] An intelligent review agent is used to review the program audio, generate review tags, and map them to the timeline file;

[0048] The review agent is also used to trigger the corresponding agent to perform targeted backtracking and partial regeneration of the smallest content unit of the corresponding type according to the review tag, and to perform partial replacement in the timeline file while keeping other smallest content units unchanged, so as to complete the automatic generation of broadcast audio programs.

[0049] The smallest content unit includes the smallest unit of program script, the smallest unit of human voice audio, the smallest unit of background music, and the smallest unit of sound effects.

[0050] The aforementioned method and system for automatic program generation based on program timelines and multi-agent collaboration can quickly output audio content with a unified style and strong adaptability by automatically generating scripts, matching timbres, background music, and sound effects according to intent information and program parameters. By splicing and synthesizing program audio according to sequential dependencies based on the program timeline, accurate audio timing and clear structure are ensured. The automatic detection and generation of review tags by the review agent accurately maps issues to the timeline. By performing targeted backtracking and partial regeneration of the smallest content units of the corresponding type, only the problematic units are replaced while other units remain unchanged, effectively avoiding full duplication of generation and reducing system resource consumption. This achieves automatic, efficient, and highly compliant production of broadcast audio programs from content creation to final output, while also reducing production costs. Attached Figure Description

[0051] Figure 1 This is an application environment diagram of an automatic program generation method based on program timeline and multi-agent collaboration in one embodiment;

[0052] Figure 2 This is a flowchart illustrating an automatic program generation method based on program timeline and multi-agent collaboration in one embodiment.

[0053] Figure 3 This is a schematic diagram illustrating the usage flow of an automatic program generation method based on program timeline and multi-agent collaboration in one embodiment.

[0054] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0056] The automatic program generation method based on program timeline and multi-agent collaboration provided in this application embodiment can be applied to, for example, Figure 1 The application environment shown. For example... Figure 1As shown, the application environment includes interconnected computer devices 110, content generation agent 120, sound generation agent 130, program synthesis agent 140, and review agent 150. Computer device 110 can obtain intent information through content generation agent 120, determine program parameters based on the intent information, and generate program script based on the program parameters; through sound generation agent 130, automatically match broadcast timbre according to program script to generate human voice audio, and determine background music and sound effects; through program synthesis agent 140, based on program timeline, splice and synthesize human voice audio, background music and sound effects according to preset order and dependency relationship to obtain program audio and timeline file; through review agent 150, review the program audio, generate review tags and map them to the timeline file; review agent 150 triggers the corresponding agent to perform targeted backtracking and partial regeneration of the corresponding type of minimum content unit according to the review tags, and performs partial replacement in the timeline file, keeping other minimum content units unchanged, to complete the automated generation of broadcast audio program; wherein, the minimum content unit includes the minimum unit of program script, the minimum unit of human voice audio, the minimum unit of background music, and the minimum unit of sound effects.

[0057] In one embodiment, such as Figure 2 As shown, an automatic program generation method based on program timeline and multi-agent collaboration is provided, including the following steps:

[0058] Step 202: Obtain intent information through the content generation agent, determine program parameters based on the intent information, and generate program script based on the program parameters.

[0059] During the copywriting stage, the content generation agent can determine user intent through multiple rounds of human-computer dialogue, obtain program topics, target duration, copywriting style, and broadcast format, and then perform parameterized settings accordingly. Specifically, the content generation agent can acquire program requirements, summarize and structure multi-source materials, and generate program copy according to the target duration and style.

[0060] In one embodiment, the provided method for automatic program generation based on program timeline and multi-agent collaboration may further include a process of generating program scripts. Specifically, the process includes: a content generation agent performing semantic parsing and intent recognition through multi-round human-computer interactive dialogue to obtain intent information including program theme, type, audience, and scene; the content generation agent configuring program parameters including target duration, script style, broadcast format, and emotional tendency based on the intent information; the content generation agent calling a large language model to summarize materials related to the program theme, forming structured content materials; and the content generation agent generating program scripts based on program parameters and structured content materials, and breaking down the program scripts into individual program script units.

[0061] The program parameters can include program topic, target duration, script style, broadcast format, paragraph structure, emotional tone, and word count constraints. The content generation agent obtains program parameters through multi-round human-computer dialogue interaction and automatically calculates the corresponding word count range based on the target duration to generate program scripts with matching word counts. Intent information can include program theme, program type, target audience, broadcast scenario, and core expression direction.

[0062] The content generation agent can use a large language model to search the internet and aggregate multi-source data on the topics input by the user, generating structured content materials. Based on the set program length, it automatically calculates the corresponding reasonable word count range, divides the overall script into logical paragraphs such as opening, body, transition, and ending according to the preset program template, allocates the word count ratio of each paragraph, and generates a complete program script that matches the theme, length, and style according to the script style parameters. It can also generate multiple versions and multiple style alternative scripts for the user to confirm and select, such as professional, warm, humorous, sharp, artistic, energetic, and minimalist. It automatically generates at least several different style versions of the script for the user to choose or confirm.

[0063] Finally, the content generation agent can break down the generated program script into multiple text units of the smallest unit according to natural paragraphs or semantic paragraphs, which serve as the basis for generating subsequent smallest content units such as human voice audio, music, sound effects, and mixing.

[0064] Step 204: The voice-generating intelligent agent automatically matches the broadcast timbre according to the program script, generates human voice audio, and determines the background music and sound effects.

[0065] In one embodiment, the provided method for automatic program generation based on program timeline and multi-agent collaboration may further include a process of generating human voice audio. The specific process includes: the voice generation agent performing semantic analysis and emotion recognition on the program script to obtain the script style, and matching the broadcast format and corresponding virtual anchor timbre according to the script style; the voice generation agent determining the speech synthesis control parameters according to the emotion recognition results of the program script, calling the speech synthesis engine, and generating human voice audio corresponding to the emotion recognition results; the voice generation agent selecting or generating background music and sound effects matching the plot according to the semantic analysis results of the program script, and breaking down the background music and sound effects into each background music minimum unit and sound effect minimum unit.

[0066] The voice generation agent can automatically match the broadcast format (unicast, multicast) and select the corresponding digital human or virtual anchor voice model based on the style and broadcast format of the text selected by the user. When the user is not satisfied with the automatic selection result, manual secondary settings are supported. Next, the voice generation agent can call the text-to-speech (TTS) engine to generate a human voice audio file with the corresponding anchor and emotional style based on the text content. Then, based on the text content and overall program style, it automatically selects appropriate background music (BGM) from the music library, or automatically generates original background music using a generative music model. Finally, it uses a large model to perform semantic analysis on the text, automatically generates sound effects elements corresponding to key plot points, and prompts the user at the corresponding positions in the text. After the user confirms, the sound effects are automatically generated and inserted.

[0067] Specifically, in this embodiment, the role of the voice generation agent can include: automatically matching the corresponding digital human or virtual anchor voice model based on the broadcast format and style of the program script, and supporting users to manually adjust the automatic matching results; performing paragraph-level semantic analysis on the program script, automatically labeling each paragraph with emotion tags, and mapping the emotion tags to control parameters recognizable by the speech synthesis engine, wherein the control parameters include at least the speech rate coefficient, fundamental frequency variation range, pause duration, and emotion intensity weight; and calling the speech synthesis engine to generate uniquely identified human voice audio for each paragraph based on the program script and control parameters, and establishing a mapping between the human voice audio and the corresponding... The program synthesis agent identifies the content relationships between program scripts; based on the emotional information of the human voice audio segments and the overall program style, it selects matching background music from the music library or calls a generative music model to automatically generate original background music, and labels each background music with its emotional attributes and applicable segment type; the program synthesis agent uses a large language model to perform semantic analysis on the program scripts, identifies environmental elements and key scenes, generates sound effect recommendation prompts at the corresponding positions in the scripts, and automatically generates the corresponding sound effect files after user confirmation; the program synthesis agent aligns the human voice audio, background music, and sound effects according to the program timeline, assigns a unique identifier to each background music and sound effect, and establishes cross-agent content relationships between human voice audio and background music and sound effects.

[0068] Step 206: Based on the program timeline, the program synthesis agent splices and synthesizes the human voice audio, background music and sound effects according to the preset order and dependency relationship to obtain the program audio and timeline files.

[0069] In one embodiment, the provided method for automatic program generation based on program timeline and multi-agent collaboration may further include a program synthesis process, which specifically includes: the program synthesis agent using the program timeline as the core data structure, and aligning each smallest content unit on the timeline according to the sequential dependency relationship of program script, human voice audio, background music, and sound effects, using a relative timestamp and anchor point alignment mechanism; the program synthesis agent calling the audio synthesis engine to perform mixing processing on the human voice audio, background music, and sound effects to obtain program audio and a structured timeline file.

[0070] The program synthesis intelligent agent can automatically splice the generated human voice, background music and sound effects according to the text order and timeline based on the audio synthesis engine (ASE); the audio processing engine can automatically mix, equalize, control the volume and dynamically process the human voice, background music and sound effects; output the final program audio file, and simultaneously generate an XML format audio timeline file for users to perform secondary fine-tuning.

[0071] That is, in this embodiment, the work of the program synthesis agent is not just a simple splicing and synthesis, but also involves automatic audio mixing, alignment and organization of the smallest content units, and generation of structured timeline files. Specifically, the program synthesis agent can first load the program timeline structure, in which the program timeline records the metadata, unique identifier, timeline position, and cross-agent content relationships of all the smallest content units in the form of a core data structure. Next, the program synthesis agent calls the audio synthesis engine, and according to the start timestamp of each smallest content unit in the program timeline, places the human voice audio on the main audio track, the background music on the background audio track, and the sound effects on the sound effects track in chronological order, and performs a timeline alignment check to ensure there are no overlapping conflicts or time gaps. According to the relative timestamp and anchor point alignment mechanism in the program timeline, the program synthesis agent precisely aligns each smallest content unit on the timeline. First, it arranges each human voice audio in the order of the script paragraphs, then aligns the corresponding background music with the human voice audio throughout, and finally positions the sound effects to the plot time points corresponding to the script and inserts them, thereby calling the audio processing engine to automatically process the spliced ​​human voice, background music, and sound effects multi-track data. Specifically, automated processing of spliced ​​data can include: enhancing voice clarity, dynamically adjusting background music volume; equalization, fade-in / fade-out, dynamic range control and noise reduction; ensuring prominent voices, natural background music, and timely sound effects without interfering with broadcasting.

[0072] After mixing and synthesis are completed, the program synthesis agent outputs a complete broadcast audio program; at the same time, it generates an XML format program timeline file that corresponds one-to-one with the audio. The program timeline file records the position, duration, type, identifier and mixing parameters of all the smallest content units, which are used for subsequent manual secondary editing, intelligent review and positioning and targeted backtracking replacement.

[0073] Step 208: The program audio is reviewed by the review agent, and review tags are generated and mapped to the timeline file.

[0074] The intelligent review agent can automatically review the synthesized program audio.

[0075] In one embodiment, the provided method for automatic program generation based on program timeline and multi-agent collaboration may further include a process of reviewing program audio. Specifically, the review agent performs multi-dimensional review checks on the program audio to obtain the results. The multi-dimensional review checks include pronunciation accuracy detection, sensitive word detection, emotional consistency detection, and audio quality detection. The review agent generates structured review tags based on the detection results. The review agent maps the review tags to the corresponding smallest content unit in the timeline file, adds review tag nodes to the timeline file, establishes a unique association between the review tags and the smallest content unit identifier, and stores the data in structured data format.

[0076] The review agent performs multi-dimensional review and inspection of program audio, including pronunciation accuracy detection, sensitive word detection, emotion consistency detection, and audio quality detection. During the review process, the detection results can be mapped to the program audio timeline and review tags can be generated. The review tags are stored in the form of structured data and used as input parameters for cross-agent task scheduling to control the backtracking scope and regeneration strategy of the corresponding agent.

[0077] The review agent performs full-scale automated inspection of the synthesized program audio. Inspection dimensions include: pronunciation accuracy detection: comparing the audio speech recognition results with the original text word by word to identify misreadings, omissions, and repetitions; sensitive content detection: screening the text and audio for sensitive words and prohibited information; emotional consistency detection: determining whether the actual broadcast emotion matches the preset emotion tags in the text; and audio quality detection: checking for abnormal volume, background noise, unbalanced mixing, and audio dropouts. The review agent can generate review tags for each inspection result, and then bind these tags by timestamp, anchor point, and smallest content unit. This precisely maps the tag positions to the corresponding paragraphs, units, and time intervals in the timeline file, making the timeline file a locationable, traceable, and schedulable problem location carrier, providing a basis for subsequent targeted backtracking and partial regeneration.

[0078] Specifically, in one embodiment, pronunciation accuracy detection converts human voice audio into recognized text using a speech recognition engine, compares the recognized text with the program script segment by segment to identify pronunciation errors, omissions, or repetitions, and maps the problem location to the corresponding smallest unit of human voice audio and the smallest unit of program script; sensitive word detection matches the program script and speech recognition results using a sensitive word database, combines contextual analysis from a large language model to identify sensitive word issues, and maps the problem location to the corresponding smallest unit of program script; emotion consistency detection analyzes the actual emotional characteristics of human voice audio using a speech emotion recognition model, compares it with preset emotion tags for each segment recorded in the timeline file, and determines an emotion mismatch problem when the emotion matching degree is lower than a set threshold, mapping the problem location to the corresponding smallest unit of human voice audio and the smallest unit of program script; audio quality detection performs volume detection, signal-to-noise ratio detection, distortion detection, and spectrum analysis on the program audio to identify abnormal volume, abnormal mixing, or abnormal sound quality issues, and maps the problem location to the smallest unit of background music and sound effects in the corresponding time interval.

[0079] Step 210: The review agent triggers the corresponding agent to perform targeted backtracking and partial regeneration of the smallest content unit of the corresponding type based on the review tag, and performs partial replacement in the timeline file while keeping other smallest content units unchanged, thus completing the automated generation of the broadcast audio program.

[0080] The smallest content unit includes the program script, the human voice audio, the background music, and the sound effects. Specifically, each smallest content unit includes at least: the program script; the human voice audio generated from the program script; the background music and sound effects units aligned with the human voice audio on the timeline; each smallest content unit has a unique identifier on the program timeline and establishes content association relationships across intelligent agents.

[0081] In one embodiment, the provided method for automatic program generation based on program timeline and multi-agent collaboration may further include a process of targeted backtracking and partial regeneration of the smallest content unit. Specifically, the process includes: the review agent parsing review tags to obtain the problem type and the smallest content unit identifier, and determining the agent to be triggered and the smallest content unit to be backtracked based on the problem type and the smallest content unit identifier; triggering the content generation agent, sound generation agent, or program synthesis agent to perform partial regeneration of the corresponding type of smallest content unit; and the program synthesis agent using a relative timestamp and anchor point alignment mechanism to partially replace the smallest content unit corresponding to the review tag in the program timeline, while maintaining the timing and content of the remaining smallest content units, thus completing the automatic generation of the broadcast audio program.

[0082] Based on the issue type in the review tags, the computer equipment automatically triggers the corresponding intelligent agent to perform targeted backtracking processing. After completing the targeted regeneration, the computer equipment only replaces the corresponding smallest content unit in the program timeline, while keeping the timeline positions of the remaining smallest content units unchanged. This achieves partial updates of the program content, avoiding the need for a complete program regeneration. The smallest content unit uses a relative timestamp and anchor point alignment mechanism in the program timeline, ensuring that unit replacement only updates the data within the corresponding time interval, without affecting the time alignment of other units.

[0083] Computer equipment can, based on the issue type and unit identifier indicated by the review tags, selectively trigger at least one of the following intelligent agents: content generation agent, sound generation agent, or program synthesis agent. This agent will then partially regenerate the corresponding smallest content unit. Based on a relative timestamp and anchor point alignment mechanism, only the issue unit will be replaced in the program timeline, while the remaining smallest content units will remain unchanged, thus completing automated program generation. In other words, the review agent parses the issue type, location information, and suggested processing method from the review tags, and determines the agent to be triggered and the smallest content unit to be backtracked based on preset issue type and agent mapping rules.

[0084] In one embodiment, when the issue type is inaccurate pronunciation or omission, the review agent only triggers the sound generation agent to obtain the text content corresponding to the review tag and regenerate the human voice audio. When the issue type in the review tag is emotion mismatch, the review agent triggers the content generation agent to rewrite the text content corresponding to the review tag, generates new text, and triggers the sound generation agent to regenerate the human voice audio based on the rewritten text. When the issue type in the review tag is sensitive words, the review agent triggers the content generation agent to rewrite the text content corresponding to the review tag for compliance, generates new program text, and triggers the sound generation agent to regenerate the human voice audio based on the rewritten text. When the issue type in the review tag is abnormal volume, abnormal mixing, or abnormal sound quality, the review agent triggers the program synthesis agent to determine the time interval corresponding to the text content of the review tag, identify the smallest content unit within the time interval, recalculate the mixing parameters, and perform local remixing.

[0085] That is, in this embodiment, when the problem type is inaccurate pronunciation or omission, only the sound generation agent is triggered to regenerate the corresponding smallest unit of human voice audio; when the problem type is emotion mismatch, the content generation agent is triggered to rewrite the corresponding smallest unit of program script with emotional expression, and the sound generation agent is triggered simultaneously to adjust the speech rate, tone and emotion parameters; when the problem type is sensitive words, only the content generation agent is triggered to rewrite the corresponding smallest unit of program script for compliance; when the problem type is abnormal volume, mixing or sound quality, only the program synthesis agent is triggered to recalculate the mixing parameters of the corresponding time interval.

[0086] This application provides a method for automatic program generation based on program timelines and multi-agent collaboration. This method automates the entire process of broadcast audio program production, from script planning to final output, significantly reducing labor costs. Through multi-agent collaboration, it improves the consistency and matching between scripts, audio, music, and sound effects. It supports the generation of multiple styles and versions of content, significantly enhancing program personalization and production efficiency. By introducing a minimum content unit based on the program timeline and a cross-agent directional backtracking mechanism, it can partially regenerate and replace the corresponding content unit when a problem is detected, significantly reducing redundant computation and improving the generation efficiency and stability of broadcast audio programs while meeting broadcast-level compliance and quality requirements. It also supports structured timeline export, balancing the needs of automated generation with meticulous manual editing.

[0087] It should be understood that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart above may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0088] In one embodiment, a program automatic generation method based on program timeline and multi-agent collaboration is applied to a real-world scenario. The planned broadcast audio program is a feature story program for broadcast on radio stations and converged media platforms, with a total program length of 15 minutes. The program format is a "narrative-style feature program," including production requirements such as emotional progression, background music changes, and enhanced environmental sound effects. The specific application process is as follows: Figure 3 As shown, it includes:

[0089] Users input the program theme "Urban Memories: The Changes of Old Streets" through the program production interface, set the program target length to 15 minutes, the overall style of the script to be "warm and artistic", the broadcast format to be "single narration + emotional broadcasting", and allow background music and sound effects to be automatically adjusted according to emotional changes;

[0090] The computer device initiates a multi-turn dialogue module, where a content-generating intelligent agent performs semantic parsing and intent recognition on the user input, confirms that the program type is "thematic narrative broadcast program," and automatically matches the corresponding program structure template;

[0091] The content generation agent calls a large language model to conduct online searches for topics, automatically obtains textual materials related to the history of the old street, urban development, and cultural memories, and performs deduplication, timeline sorting, and sentiment tagging on multi-source data;

[0092] Based on the theme program template, the content generation AI automatically breaks down the copy into the following logical paragraphs: 1) Opening introduction paragraph, used to create an emotional atmosphere; 2) Historical background paragraph, used to introduce the formation and development of the old street; 3) Characters and stories paragraph, used to describe specific characters or life scenes; 4) Changes and comparison paragraph, used to reflect the changes in urban development; 5) Emotional sublimation and summary paragraph, used to conclude the story.

[0093] The content generation AI automatically calculates the target word count ratio for each segment based on the 15-minute program length and generates a complete script. At the same time, based on style parameters such as "warmth" and "artistic", it automatically generates at least several different versions of the script with different wording and expressions for users to choose from or fine-tune. The smallest content unit is divided into natural paragraphs of the script, and each natural paragraph corresponds to a human voice audio unit and its aligned music and sound effects unit in the timeline.

[0094] After the user confirms the target text, the voice generation AI automatically selects a digital anchor voice model with strong emotional expression as the default broadcasting role based on the program type and text style.

[0095] The voice generation agent performs paragraph-level semantic analysis on the text and automatically labels different paragraphs with emotion tags, including but not limited to "nostalgia, calm, narrative, sentiment, sublimation"; and when calling the speech synthesis engine, the emotion tags are further mapped to control parameters that the speech synthesis engine can recognize, including but not limited to speech rate coefficient, fundamental frequency variation range, pause duration and emotion intensity weight.

[0096] The generative music model generates multiple original background music tracks based on the overall style of the program, and labels their emotional attributes and applicable segment types; the sound generation agent automatically selects and switches different background music tracks on the program timeline based on the emotional changes of the segments.

[0097] Meanwhile, the sound generation intelligent agent identifies environmental elements and key scenes based on the text content, automatically recommends environmental sound effects or ambient sound effects in places such as "old street in the early morning", "people coming and going" and "changing times", and makes visual markings in the corresponding positions of the text for users to confirm and automatically generate and insert.

[0098] The program synthesis agent calls the audio synthesis engine to splice together the human voice audio, background music and sound effect files generated by segment according to the preset timeline order; and through the audio processing engine, it automatically mixes, equalizes, fades in and out and controls the dynamic range of different audio tracks to make the human voice clear and the background music naturally connected; finally, it generates a complete theme program audio file and simultaneously outputs the corresponding XML format timeline file for users to perform subsequent fine editing.

[0099] The intelligent review agent first performs speech recognition on the human voice audio and then compares the recognized text with the original text segment by segment to detect whether there are pronunciation errors, omissions, or repetitions.

[0100] Subsequently, the intelligent review agent performs sensitive word detection, emotion consistency detection, and audio quality detection on the synthesized program audio. When a problem is detected, the problem location is automatically marked on the timeline, and the problem is fed back to the corresponding processing stage, where the relevant intelligent agent automatically regenerates the corresponding content until the output of the special program audio that meets the broadcast standards is obtained.

[0101] For example, when the intelligent review agent detects that the emotional intensity of the human voice audio is lower than the set threshold for the corresponding paragraph within a certain time period, the system automatically determines that it is an emotional mismatch problem, and triggers the content generation agent to rewrite the text of the paragraph with enhanced emotion, while triggering the sound generation agent to regenerate the corresponding human voice audio unit.

[0102] In this embodiment, the program timeline is used as the core data structure. The program is divided into multiple independently generated and replaceable minimum content units. Based on these minimum content units, targeted generation, review, and retrospective optimization across intelligent agents are achieved, thereby completing the end-to-end automated production of broadcast audio programs. The generation of the minimum content units follows the sequential dependency relationship of "text unit → vocal audio unit → music and sound effects unit → mixing unit". The generation of subsequent units uses the output results of their predecessor units as input conditions, achieving high-efficiency, high-consistency, and high-compliance program production.

[0103] In one embodiment, such as Figure 1 As shown, a system for automatic program generation based on program timeline and multi-agent collaboration is provided, including:

[0104] Content generation agent 120 is used to acquire intent information, determine program parameters based on intent information, and generate program scripts based on program parameters.

[0105] The sound generation agent 130 is used to automatically match the broadcast timbre according to the program script, generate human voice audio, and determine the background music and sound effects.

[0106] The program synthesis agent 140 is used to splice and synthesize human voice audio, background music and sound effects according to a preset order and dependency relationship based on the program timeline to obtain program audio and timeline files.

[0107] The review agent 150 is used to review program audio, generate review tags, and map them to the timeline file;

[0108] The review agent 150 is also used to trigger the corresponding agent to perform targeted backtracking and partial regeneration of the smallest content unit of the corresponding type based on the review tag, and to perform partial replacement in the timeline file while keeping other smallest content units unchanged, so as to complete the automated generation of broadcast audio programs.

[0109] The smallest content unit includes the program script, the human voice audio, the background music, and the sound effects.

[0110] In one embodiment, the content generation agent 120 is also used to perform semantic parsing and intent recognition through multi-round human-computer interactive dialogue to obtain intent information including program theme, type, audience, and scene; to complete the configuration of program parameters including target duration, copywriting style, broadcasting format, and emotional tendency based on the intent information; to call a large language model to summarize the program theme and form structured content materials; and to generate program copy based on the program parameters and the structured content materials, and to break down the program copy into the smallest unit of each program copy.

[0111] In one embodiment, the voice generation agent 130 is further used to perform semantic analysis and emotion recognition on the program script to obtain the script style, and match the broadcast format and corresponding virtual anchor voice according to the script style; determine the speech synthesis control parameters according to the emotion recognition result of the program script, call the speech synthesis engine, and generate human voice audio corresponding to the emotion recognition result; select or generate background music and sound effects that match the plot according to the semantic analysis result of the program script, and break down the background music and sound effects into each background music minimum unit and sound effect minimum unit.

[0112] In one embodiment, the program synthesis agent 140 is also used to align each smallest content unit on the timeline according to the program timeline as the core data structure, based on the sequential dependency relationship of program script, human voice audio, background music and sound effects, using a relative timestamp and anchor point alignment mechanism; and to call the audio synthesis engine to perform mixing processing on human voice audio, background music and sound effects to obtain program audio and structured timeline files.

[0113] In one embodiment, the review agent 150 is also used to perform multi-dimensional review and detection on the program audio to obtain the detection results; the multi-dimensional review and detection includes pronunciation accuracy detection, sensitive word detection, emotion consistency detection and audio quality detection; generate structured review tags based on the detection results; map the review tags to the corresponding smallest content unit in the timeline file, add review tag nodes in the timeline file, establish the association between the review tags and the unique identifier of the smallest content unit, and store them in the form of structured data.

[0114] In one embodiment, pronunciation accuracy detection converts human voice audio into recognized text using a speech recognition engine, compares the recognized text with the program script segment by segment to identify pronunciation errors, omissions, or repetitions, and maps the problem location to the corresponding smallest unit of human voice audio and the smallest unit of program script. Sensitive word detection matches the program script and speech recognition results using a sensitive word database, combines contextual analysis with a large language model to identify sensitive word issues, and maps the problem location to the corresponding smallest unit of program script. Emotion consistency detection analyzes the actual emotional characteristics of human voice audio using a speech emotion recognition model, compares them with preset emotion labels for each segment recorded in the timeline file, and determines an emotion mismatch problem when the emotion matching degree is lower than a set threshold, mapping the problem location to the corresponding smallest unit of human voice audio and the smallest unit of program script. Audio quality detection performs volume detection, signal-to-noise ratio detection, distortion detection, and spectrum analysis on the program audio to identify abnormal volume, abnormal mixing, or abnormal sound quality issues, and maps the problem location to the smallest unit of background music and sound effects in the corresponding time interval.

[0115] In one embodiment, the review agent 150 is further configured to parse the review tag to obtain the issue type and the minimum content unit identifier, and determine the agent to be triggered and the minimum content unit to be backtracked based on the issue type and the minimum content unit identifier; trigger the content generation agent 120, the sound generation agent 130 or the program synthesis agent 140 to partially regenerate the minimum content unit of the corresponding type; the program synthesis agent 140 is further configured to use a relative timestamp and anchor point alignment mechanism to partially replace the minimum content unit corresponding to the review tag in the program timeline, while keeping the timing and content of the remaining minimum content units unchanged, thereby completing the automated generation of the broadcast audio program.

[0116] In one embodiment, when the issue type is inaccurate pronunciation or omission, the review agent 150 only triggers the sound generation agent 130 to obtain the text content corresponding to the review tag and regenerate the human voice audio. When the issue type in the review tag is emotion mismatch, the review agent 150 triggers the content generation agent 120 to rewrite the text content corresponding to the review tag, generate new text, and triggers the sound generation agent 130 to regenerate the human voice audio based on the rewritten text. When the issue type in the review tag is sensitive words, the review agent 150 triggers the content generation agent 120 to rewrite the text content corresponding to the review tag for compliance, generate new program text, and triggers the sound generation agent 130 to regenerate the human voice audio based on the rewritten text. When the issue type in the review tag is abnormal volume, abnormal mixing, or abnormal sound quality, the review agent 150 triggers the program synthesis agent 140 to determine the time interval corresponding to the text content of the review tag, identify the smallest content unit within the time interval, recalculate the mixing parameters, and perform local remixing.

[0117] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When executed by the processor, the computer program implements an automatic program generation method based on a program timeline and multi-agent collaboration. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0118] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0119] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of an automatic program generation method based on program timeline and multi-agent collaboration.

[0120] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program being executed by a processor to implement the steps of an automatic program generation method based on a program timeline and multi-agent collaboration.

[0121] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0122] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0123] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A program automatic generation method based on program timeline and multi-agent cooperation, characterized in that, The method includes: The content generation agent obtains intent information, determines program parameters based on the intent information, and generates program scripts based on the program parameters. The voice-generating intelligent agent automatically matches the broadcast timbre according to the program script, generates human voice audio, and determines background music and sound effects. Based on the program timeline, the program synthesis agent splices and synthesizes the human voice audio, background music and sound effects according to a preset order and dependency relationship to obtain the program audio and timeline files. The program audio is reviewed by an intelligent review agent, which generates review tags and maps them to the timeline file. The review agent triggers the corresponding agent to perform targeted backtracking and partial regeneration of the smallest content unit of the corresponding type according to the review tag, and performs partial replacement in the timeline file while keeping other smallest content units unchanged, thus completing the automated generation of broadcast audio programs; The smallest content unit includes the smallest unit of program script, the smallest unit of human voice audio, the smallest unit of background music, and the smallest unit of sound effects. 2.The program automatic generation method based on program timeline and multi-agent collaboration according to claim 1, characterized in that, The process involves acquiring intent information through a content generation agent, determining program parameters based on the intent information, and generating program scripts based on the program parameters, including: The content-generating intelligent agent performs semantic parsing and intent recognition through multi-round human-computer interactive dialogue to obtain intent information including program theme, type, audience, and scene; The content generation agent configures program parameters, including target duration, copywriting style, broadcasting format, and emotional tone, based on the intent information. The content generation agent calls a large language model to summarize materials for the program theme and form structured content materials; The content generation agent generates program scripts based on the program parameters and the structured content materials, and then breaks down the program scripts into the smallest units of program scripts. 3.The program automatic generation method based on program timeline and multi-agent collaboration according to claim 1, characterized in that, The voice-generating intelligent agent automatically matches the broadcast timbre to the program script, generates human voice audio, and determines background music and sound effects, including: The voice generation agent performs semantic analysis and emotion recognition on the program script to obtain the script style, and matches the broadcast format and corresponding virtual anchor voice according to the script style; The voice generation agent determines the speech synthesis control parameters based on the emotion recognition results of the program script, calls the speech synthesis engine, and generates human voice audio corresponding to the emotion recognition results. The sound generation agent selects or generates background music and sound effects that match the plot based on the semantic analysis results of the program script, and breaks down the background music and sound effects into each smallest unit of background music and the smallest unit of sound effects.

4. The method of claim 1, wherein the program timeline and multi-agent collaboration based automatic program generation method is characterized by, Based on the program timeline, the program synthesis agent splices and synthesizes the human voice audio, background music, and sound effects according to a preset order and dependency relationship to obtain program audio and timeline files, including: The program synthesis intelligent agent uses the program timeline as the core data structure. According to the sequential dependency relationship of program script, human voice audio, background music and sound effects, it uses a relative timestamp and anchor point alignment mechanism to align each of the smallest content units on the timeline. The program synthesis agent calls the audio synthesis engine to mix the human voice audio, background music and sound effects to obtain program audio and structured timeline files.

5. The method of claim 1, wherein the program timeline and multi-agent collaboration based automatic program generation method is characterized by, The program audio is reviewed by an intelligent review agent, review tags are generated and mapped to the timeline file, including: The review agent performs multi-dimensional review and inspection on the program audio to obtain the inspection results; the multi-dimensional review and inspection includes pronunciation accuracy detection, sensitive word detection, emotion consistency detection, and audio quality detection; The auditing agent generates structured auditing tags based on the detection results; The review agent maps the review tags to the corresponding smallest content unit in the timeline file, adds review tag nodes to the timeline file, establishes a unique association between the review tags and the smallest content unit, and stores it in the form of structured data.

6. The method of claim 5, wherein, The review agent performs multi-dimensional review and detection on the program audio, and obtains the detection results, including: The pronunciation accuracy detection uses a speech recognition engine to convert human voice audio into recognized text, compares the recognized text with the program script segment by segment, identifies pronunciation errors, omissions or repetitions, and maps the problem location to the corresponding smallest unit of human voice audio and the smallest unit of program script. The sensitive word detection matches the program text and speech recognition results using a sensitive word database, and combines the contextual analysis of a large language model to identify sensitive word issues and map the problem location to the corresponding smallest unit of the program text. The emotion consistency detection analyzes the actual emotion characteristics of human voice audio through a voice emotion recognition model and compares them with the preset emotion tags of each segment recorded in the timeline file. When the emotion matching degree is lower than the set threshold, it is determined to be an emotion mismatch problem, and the problem location is mapped to the corresponding smallest unit of human voice audio and the smallest unit of program script. The audio quality detection performs volume detection, signal-to-noise ratio detection, distortion detection, and spectrum analysis on the program audio to identify problems such as abnormal volume, abnormal mixing, or abnormal sound quality, and maps the problem location to the smallest unit of background music and the smallest unit of sound effects in the corresponding time interval.

7. The method of claim 1, wherein the program timeline and multi-agent collaboration based automatic program generation method is characterized by, The review agent triggers the corresponding agent to perform targeted backtracking and partial regeneration of the corresponding type of minimum content unit based on the review tag, and performs partial replacement in the timeline file while keeping other minimum content units unchanged, thus completing the automated generation of broadcast audio programs, including: The review agent parses the review tag to obtain the problem type and the smallest content unit identifier, and determines the agent to be triggered and the smallest content unit to be traced back based on the problem type and the smallest content unit identifier. Trigger the content generation agent, sound generation agent, or program synthesis agent to locally regenerate the smallest content unit of the corresponding type; The program synthesis agent uses a relative timestamp and anchor point alignment mechanism to partially replace the smallest content unit corresponding to the review tag in the program timeline, while keeping the timing and content of the remaining smallest content units unchanged, thus completing the automated generation of broadcast audio programs. 8.The program automatic generation method based on program timeline and multi-agent collaboration according to claim 7, characterized in that, The triggering agent for content generation, sound generation, or program synthesis performs local regeneration of the smallest content unit of the corresponding type, including: When the problem type is inaccurate pronunciation or omission, the review agent only triggers the sound generation agent, which obtains the text content corresponding to the review tag and regenerates the human voice audio. When the issue type in the review tag is "emotion mismatch", the review agent triggers the content generation agent to rewrite the text content corresponding to the review tag, generate new text, and triggers the sound generation agent to regenerate human voice audio based on the rewritten text. When the issue type in the review tag is a sensitive word, the review agent triggers the content generation agent to rewrite the text content corresponding to the review tag in a compliant manner, generates a new program script, and triggers the sound generation agent to regenerate human voice audio based on the rewritten script. When the problem type in the review label is abnormal volume, abnormal mixing, or abnormal sound quality, the review agent triggers the program synthesis agent to determine the time interval corresponding to the text content of the review label, identify the smallest content unit within the time interval, recalculate the mixing parameters, and perform local remixing.

9. A program automatic generation system based on program timeline and multi-agent collaboration, characterized in that, The system includes: A content generation agent is used to acquire intent information, determine program parameters based on the intent information, and generate program scripts based on the program parameters. A voice generation intelligent agent is used to automatically match the broadcast timbre according to the program script, generate human voice audio, and determine background music and sound effects; The program synthesis intelligent agent is used to splice and synthesize the human voice audio, background music and sound effects according to a preset order dependency relationship based on the program timeline to obtain program audio and timeline files; An intelligent review agent is used to review the program audio, generate review tags, and map them to the timeline file; The review agent is also used to trigger the corresponding agent to perform targeted backtracking and partial regeneration of the smallest content unit of the corresponding type according to the review tag, and to perform partial replacement in the timeline file while keeping other smallest content units unchanged, so as to complete the automatic generation of broadcast audio programs. The smallest content unit includes the smallest unit of program script, the smallest unit of human voice audio, the smallest unit of background music, and the smallest unit of sound effects.

Images