PEPETII/danmuai

What it solves

DanmuAI creates an "AI audience" for your computer screen. It solves the problem of having static content (like livestreams, games, or recordings) by adding real-time, AI-generated "danmu" (scrolling comments) that react to what is happening on screen, making the viewing experience more interactive and lively.

How it works

The application follows a specific pipeline to generate reactions:

1. Screen Capture: It takes screenshots of a designated monitor at set intervals.
2. Visual Analysis: These images are compressed in memory and sent to a vision-capable AI model to understand the current content.
3. Comment Generation: The AI generates comments based on the screen content and a configurable "persona" (personality).
4. Display: The comments are displayed on a transparent, always-on-top overlay on the screen.
5. Audio: Optional Text-to-Speech (TTS) converts the comments into spoken audio using various providers.

Who it’s for

• Streamers and content creators who want an automated AI commentator for their broadcasts.
• Gamers who want a reactive AI companion while playing.
• Users who want to add a dynamic, social-like comment layer to any desktop activity.

Highlights

• Customizable Personas: Configure the AI's personality, model platform, and tone to match your preference.
• Transparent Overlay: A dedicated scrolling comment layer that stays on top of other windows without needing the web console open.
• Multi-Monitor Support: Ability to capture screenshots from one monitor and display the comments on another.
• TTS Integration: Supports multiple AI voice providers (e.g., MiMo, Qwen3, MiniMax, Doubao V3) with voice caching and previews.
• Optional Extensions: Includes support for microphone input, Live2D virtual avatars, and knowledge bases.