Demo Online: https://demo.byteplus.com/rtc/solution/aigc

• In the AIGC conversation scenario, the Volcengine AIGC - RTC Server cloud service provides an end - to - end AIGC capability link based on streaming voice by integrating RTC audio and video stream processing, ASR voice recognition, large - model interface call integration, and TTS voice generation capabilities.
• Users only need to call the standard - based OpenAPI interfaces to configure the required ASR, LLM, and TTS types and parameters. The Volcengine cloud computing service is responsible for edge user access, cloud resource scheduling, audio and video stream compression, text - to - voice conversion processing, and data subscription and transmission. This simplifies the development process, allowing developers to focus more on the training and debugging of the core capabilities of large models, thus rapidly promoting the innovation of AIGC product applications.
• At the same time, Volcengine RTC has mature technologies such as audio 3A processing and video processing, as well as large - scale audio and video chat capabilities. It can enable AIGC products to more conveniently support scenarios such as multimodal interaction and multi - person interaction, maintaining the naturalness and efficiency of interaction.

This demo supports two operation modes with different configuration requirements:

Allows independent configuration of ASR, LLM, and TTS components with multiple provider options.

You need to fill the following fields into VOICE_CHAT_MODE in ./Server/sensitive.js according to the ASR/LLM/TTS vendor you use.

Besides, if using BytePlusArk as LLM Module, you need to fill the model endpoint ID in ArkModel (File: src/config/voiceChat/llm.ts), which could be gain from BytePlus Ark Console.

VOICE_CHAT_MODE
├── ASRConfig (Speech Recognition)
│   ├── amazon
│   │   ├── ID (Access Key ID)
│   │   └── Secret (Secret Access Key)
│   ├── google
│   │   └── CredentialsJSON (Service Account JSON)
│   └── BytePlus
│       ├── AppId (BytePlus ASR App ID)
│       ├── AccessToken (BytePlus ASR Access Token)
│       └── ApiResourceId (Service Plan Type)

│
├── TTSConfig (Text-to-Speech)
│   ├── byteplus_Bidirectional_streaming
│   │   └── app
│   │       ├── appid (BytePlus TTS App ID)
│   │       └── token (BytePlus TTS Access Token)
│   ├── amazon
│   │   ├── ID (Access Key ID)
│   │   └── Secret (Secret Access Key)
│   └── openai
│       ├── URL (API Endpoint)
│       └── APIKey (OpenAI API Key)
│
└── LLMConfig (Large Language Model)
│   ├── CustomLLM (OpenAI Compatible)
│   │   ├── Url (API Endpoint)
│   │   └── APIKey (API Key)
│   └── BytePlusArk
│       └── APIKey (BytePlus Ark API Key)
│ 
└── AvatarConfig
    ├── AvatarAppID
    └── AvatarToken

Uses OpenAI's integrated ASR+TTS solution with only LLM being configurable.

You need to fill the OpenAI token into REALTIME_API_MODE in ./Server/sensitive.js.

REALTIME_API_MODE
└── LLMConfig (OpenAI Realtime API)
    └── Token (OpenAI API Key)

• Node Version: 16.0+

1. Two terminals are required to start the server and the front-end page respectively.

2. RTC Basic Configuration (src/config/config.ts)

   • AppId: Your BytePlus RTC App ID (required)
   • RoomId: Auto-generated UUID or custom room ID (optional)
   • UserId: Auto-generated UUID or custom user ID (optional)
   • Token: Token generated in Byteplus Console or leave it undefined for token auto-generation, demo will invoke api(defined in ./Server/app.js) to generate token, which require your RTC_APP_KEY in ./Server/sensitive.js.

3. Server Configuration (Server/sensitive.js)

   • RTC Basic Configuration: Configure RTC_APP_KEY if you want to auto generate token in server.
   • API Provider Credentials:
     • Flexible Mode:
       • Configure ASR, TTS, LLM, and Avatar ( if enabled ) provider credentials as shown in the configuration tree above.
       • If using BytePlusArk as LLM Module, you need to fill the model endpoint ID in ArkModel (File: src/config/voiceChat/llm.ts), which could be gain from BytePlus Ark Console.
     • Realtime Mode: Only requires OpenAI APIKey in LLMConfig.Token.

Refer to the configuration tree structure above for the complete list of required fields.

Please note that both the server and the web need to be started. The steps are as follows:

Enter the project root directory

cd Server
yarn

node app.js

or

yarn dev

Enter the project root directory

yarn

For Flexible Mode (customizable ASR/LLM/TTS):

yarn dev:flexible

For Realtime Mode (OpenAI integrated solution):

yarn dev:realtime

If you encounter issues beyond those listed above, please feel free to contact us for feedback.

• Overview
• Demo Experience
• Solution Integration

Refer to OpenAPI Updates for updates related to Real-time Conversational AI.

• 2025-11-06

  • Support BytePlus Avatar

• 2025-08-04

  • Upgrade SDK version to 4.67.2
  • Support BytePlus ASR（Speech-to-Text (ASR) - Streaming）
  • fix AI Settings UI

• 2025-07-08

  • Upgrade SDK version to 4.66.20

• 2025-06-16

  • Updated Demo guide and fixed some parameter errors.
  • Modify the Endpoint ID of Byteplus Ark Model. which needs to be filled in by the user.

• 2025-06-06

  • Initial Release