AI Chatter is deprecated: this feature is a proof-of-concept, is no longer actively maintained, and will be removed in a future release. Its more capable successor LAPLACE Chatterbox offers richer character setup and audience danmaku context — migrate to it
LAPLACE Cap's AI Chatter feeds the rolling text of an STT live transcript into an LLM, generates audience-style danmaku of up to 40 characters, and can send them to a bilibili live room automatically
Features
- Built on any OpenAI-compatible API, so it works with OpenAI, Anthropic, OpenRouter, and other major LLM services
- A customizable system prompt for tuning the AI's character, interaction style, and how often it speaks
- Multiple prompt presets for switching quickly between live-stream scenarios
- Optional LAPLACE Event Bridge integration, which reads real audience danmaku from the live room as context so generated danmaku blend into the current mood
- A configurable speaking interval that controls how often AI danmaku is sent
- A bilibili token from LAPLACE Login Sync, which enables automatic sending
Usage
Prerequisites
Fill in the following on the General settings page:
| Field | Description |
|---|---|
| LLM API Endpoint | LLM API address; defaults to Anthropic, replaceable with any OpenAI-compatible endpoint |
| LLM API Key | The API key for that service |
| LLM Model | The model name to use, such as claude-sonnet-4-6 or gpt-4o |
| Login Sync Token | The bilibili sign-in token used to send danmaku automatically, obtained via LAPLACE Login Sync |
Configuring AI Chatter
Switch to the AI Chatter tab:
- Enter the target live room ID
- Adjust the AI's character and interaction style in the system prompt editor
- Tune Max Context Characters and Speaking Interval as needed
Starting to Chat
- Make sure live transcription is running on the STT tab
- Click Start in AI Chatter; the AI reads the transcript continuously and generates danmaku on a cycle
- Generated danmaku go to the configured live room once you confirm them, or directly in automatic mode
System Prompt
The system prompt determines how the AI behaves. The default preset holds these customizable blocks:
## Current Stream Topic:
（当前直播的内容主题，例如游戏名称或活动类型）
## Your Character Setup:
（AI 扮演的观众角色设定，包括性格、语气、常用语等）
## Current Streamer Character:
（主播的性格描述，让 AI 了解互动对象）
## Current Audience Character:
（目标观众群体特征，使 AI 的发言风格与弹幕区整体氛围一致）The more specific the prompt, the more natural the generated danmaku. Spell out the streamer's nickname, the names the audience uses, and the running jokes in chat
Connecting LAPLACE Event Bridge
With Event Bridge connected, generation also draws on the real audience danmaku in the live room, keeping the output close to the topics and mood of the current chat
Under Event Bridge in AI Chatter:
- Turn on Enable Event Bridge
- Enter the Event Bridge WebSocket address (defaults to
wss://event-fetcher.laplace.cn/) - Enter the auth token and the room ID to subscribe to
For deploying and using Event Bridge, see the LAPLACE Event Bridge documentation
Notes
- AI Chatter calls the LLM API continuously. Set the speaking interval against your provider's pricing to avoid excessive costs
Last updated on September 19, 2026