Voice recognition (voice typing) in Thunderbird

Prefer to watch? Set up private, on-device voice recognition:

Watch: local Whisper

Voice typing lets you dictate into the atbridge AI Chat instead of typing — press the mic, speak, and your words appear in the composer as text. It runs on whatever transcription model you connect, including a built-in Whisper that runs entirely on your own machine (no API key, nothing leaves your computer).

Note

Voice typing in the chat composer is a Pro feature. Transcribing an attached or recorded audio file (see Audio notes) is available on Personal and Pro.

1. Choose a voice model

Open the atbridge settings (the @b button → Settings, or Open the extension settings) and go to Chat → Preferences → AI for voice.

1. Pick a transcription provider and model. Any provider that supports speech-to-text works:

   • OpenAI — whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe
   • Groq — whisper-large-v3, whisper-large-v3-turbo
   • Google Gemini — gemini-2.5-flash / -pro
   • Any OpenAI-compatible endpoint — point it at your own URL
   • Built-in local Whisper — “Built-in local Whisper (private — no key)”. Runs on-device; nothing is sent to any provider.

2. Built-in Whisper: download a model. When you pick the built-in engine, a Download model button appears. Choose a size — tiny (~75 MB), base (~148 MB), or small (~466 MB); larger is more accurate, smaller is faster. The download runs once, then transcription is fully offline.

3. (Optional) “Use always”. Turn this on to make this voice model the default for every transcription (voice typing and audio notes).

2. Set the language you speak

Under AI for voice set Voice input language (what you speak). Leave it on Automatic (match my language) to follow your interface language, or pin a specific language. Pinning helps short clips — auto-detection can drift on a few words, so if a quick phrase comes back in the wrong language, set the language explicitly.

3. Dictate in the AI Chat

1. Open the atbridge AI Chat in Thunderbird.

2. Click the microphone button in the composer (“Voice typing — dictate your message”). The first time, your browser/Thunderbird asks for microphone permission — allow it.

3. Speak, then stop. Your speech is transcribed and inserted into the composer, where you can edit it and send like any other message.

Tip

No mic button? It’s Pro-only — check your plan. If it errors before the mic opens, no voice model is configured yet — finish step 1 above.

Privacy

Recording and playback happen on your machine. The audio is sent only to the transcription provider you chose — and with the built-in local Whisper, not even that: it never leaves your computer. atbridge never routes your audio through its own servers.

Related

• Audio notes & transcription — record or attach a clip in atbridge Notes and transcribe it.
• Use the atbridge AI Chat — where voice typing lives.
• Set up the extension — install and first-run setup.