GeminiVRM

English

日本語

English

日本語

Appearance

Getting Started ​

Prerequisites ​

  • Node.js 20 or later
  • A Gemini API key from Google AI Studio
  • A browser with microphone and WebGL support if you want the full chat + avatar flow
  • A Google OAuth web client ID if you want to use the optional YouTube Live relay

Install Dependencies ​

bash
npm install

Run The App ​

bash
npm run dev -- --hostname 127.0.0.1 --port 3100

Open http://127.0.0.1:3100, paste your Gemini API key, and press Start.

GeminiVRM is currently tuned for gemini-3.1-flash-live-preview. Keep NEXT_PUBLIC_GEMINI_LIVE_MODEL on that value unless you are intentionally testing another Live model. Automatic fallback to older preview models is disabled.

Run The App And Docs Together ​

bash
npm run dev:all

This starts:

  • the app at http://127.0.0.1:3100
  • the VitePress docs at http://127.0.0.1:4173

First Successful Session Checklist ​

  1. Confirm the default Kiyoka.vrm model loads.
  2. Enter a Gemini API key.
  3. Send a short prompt from the message box.
  4. Wait for the assistant transcript and streamed audio playback.
  5. Open Settings to tune the model, voice, system prompt, and motion preset.

Optional Podcast Mode Quick Start ​

If you want to verify the new dual-host flow right away:

  1. Open Settings and switch Conversation mode to Podcast mode.
  2. Wait for both Kiyoka and Yukito viewers to finish loading.
  3. Open Podcast settings if you want to change the max loop count or assign different prebuilt voices per host.
  4. Type a short topic and send it.
  5. Confirm the stage alternates speakers and stops automatically when the configured turn count is reached.

Optional YouTube Relay Setup ​

If you want live chat comments to flow into Gemini:

  1. Create a Google OAuth web client ID that allows the exact origin you use for the app.
  2. Set NEXT_PUBLIC_GOOGLE_CLIENT_ID or paste the client ID into Settings -> Streaming -> YouTube relay.
  3. Sign in with Google, refresh the broadcast list, and pick the active or upcoming stream you want to monitor.
  4. Enable relay mode, then optionally enable auto-reply once the stream is ready.
  5. The browser restores the saved client ID and short-lived access token from local storage until you sign out or the token expires.

For the full setup, relay behavior, and troubleshooting flow, see the YouTube Relay Guide.

Useful Environment Variables ​

  • NEXT_PUBLIC_GEMINI_API_KEY
  • NEXT_PUBLIC_GOOGLE_CLIENT_ID
  • NEXT_PUBLIC_GEMINI_LIVE_MODEL
  • NEXT_PUBLIC_GEMINI_LIVE_VOICE
  • BASE_PATH
  • NEXT_PUBLIC_DOCS_URL