# AITuber OnAir

AITuber OnAir is a browser-based AI avatar app for direct chat and live-streaming workflows. It lets a user run an AI character, connect to audience comments, generate speech, and display an avatar without installing a dedicated desktop app.

Primary site: https://aituberonair.com/
Human manual: https://aituberonair.com/manual/en/
Japanese manual: https://aituberonair.com/manual/ja/
Bug reports: https://github.com/shinshin86/aituber-onair/issues

## What the app can do

- Run an AI avatar in a web browser.
- Chat directly with the avatar from the browser UI.
- Read and respond to YouTube Live comments.
- Read and respond to Twitch chat.
- Use WebSocket chat and initiative messages for external integrations.
- Generate spoken replies with local or cloud voice engines.
- Display avatars as VRM, Live2D, image/video media avatars, PNGTuber avatars, or no avatar.
- Use presets to switch character, model, voice, streaming, and UI settings.
- Store chat history locally and import/export app data.
- Use optional vision, FAQ, announcement, NG filter, and repeated-response prevention features.

## Main setup flow

1. Open https://aituberonair.com/ in Google Chrome or another supported browser.
2. Open settings from the app menu.
3. Configure an AI chat provider and API key, unless using a local or browser-only provider.
4. Configure avatar type and avatar assets.
5. Configure avatar and listener voice engines.
6. For streaming, configure YouTube Live, Twitch, or WebSocket chat settings.
7. For external broadcasting, capture the browser window in OBS and route audio as needed.
8. Save a preset when the configuration should be reused.

## AI chat providers

Supported provider families include OpenAI, Gemini, Gemini Nano, Claude, Z.ai, xAI, Kimi, DeepSeek, Mistral, OpenRouter, and OpenAI-compatible APIs. Availability, model names, pricing, and API-key requirements depend on each provider.

## Voice engines

Supported voice engine families include VOICEVOX, AivisSpeech, OpenAI TTS, Gemini TTS, xAI TTS, Unreal Speech, ElevenLabs, Inworld, Piper Plus, OpenAI Compatible TTS, VOICEPEAK, Aivis Cloud, and MiniMax Audio.

Local engines such as VOICEVOX, AivisSpeech, VOICEPEAK, and some compatible endpoints may require a local server and an allowed origin of `https://aituberonair.com`.

## Avatar types

- VRM: 3D avatar model, commonly created with VRoid Studio.
- Live2D: Live2D model folder containing a `.model3.json` file.
- Media avatar: image or video file.
- PNGTuber: image set for avatar states such as mouth open and mouth closed.
- None: hide the avatar.

Do not assume Inochi2D is available in the released public manual yet.

## Data and privacy notes

- API keys and app settings are stored in the browser environment used by the app.
- Chat history and imported avatar/media assets are managed locally by the browser.
- When a cloud AI or voice provider is selected, prompts, comments, and generated text may be sent to that provider according to its terms.
- The app is a frontend web app; external streaming still depends on tools such as OBS, platform settings, and local audio routing.

## Useful pages

- Manual, Japanese: https://aituberonair.com/manual/ja/
- Manual, English: https://aituberonair.com/manual/en/
- Manual, Korean: https://aituberonair.com/manual/ko/
- Manual, Simplified Chinese: https://aituberonair.com/manual/zh/
- Manual, Traditional Chinese: https://aituberonair.com/manual/zh-TW/
- Manual, Thai: https://aituberonair.com/manual/th/