Muse Voice Transcribe

Muse Voice Transcribe is Meta’s speech model, offered through the Meta Model API. In Ayumi it is a third cloud engine next to Gemini and Gemini Transcribe: it writes a transcript — no prompt — with speaker labels, turn timestamps, language hints and keyword biasing. You bring your own API key.

Tip

Ayumi transcribes on-device out of the box. Reach for Muse when you want speaker-separated meeting notes or its language coverage, and compare it with Gemini Transcribe on your own recordings.

Getting a Meta Model API Key

1. Sign in to the Meta developer dashboard
2. Create an API key for the Model API
3. Set up billing — the transcribe endpoint refuses requests until a payment method is on the account
4. Copy the key

Caution

Keep your API key private. Never share it publicly.

Configuring in Ayumi

1. Open Transcription Settings
2. Paste the key under Meta
3. Star the built-in Muse Voice Transcribe profile to make it the default, or pick it per recording

The key under Meta is the provider’s base key. Every Muse profile uses it unless the profile sets an Override API Key of its own. Keys are stored securely in your device’s Keychain.

Options

Option	What it does
Label Speakers	One paragraph per speaker turn (**Speaker 1:** …). Muse separates 20 or more speakers.
Turn Timestamps	A [mm:ss–mm:ss] range on each turn. Word-level times are not available from this model.
Languages	Hints for the model, chosen from the 25 languages it supports. Leave it on Automatic to detect.
Keywords	Names, jargon and product names the model might mishear, from a newline-separated Markdown file (for example in iCloud Drive), terms typed into the profile, or both. The same file works for Gemini Transcribe.

Keywords bias recognition but do not guarantee an exact spelling. Unlike Gemini Transcribe, keywords can be combined with speaker labels.

Limits

• Recordings up to 10 minutes. Longer recordings are refused before anything is sent; use another engine for them.
• Ayumi converts the recording to 16 kHz mono PCM on the device before upload, so a ten-minute recording stays under the API’s 32 MB request cap.
• Pricing is per audio minute; see Meta’s pricing page.

Data Privacy

When using a Muse profile:

• Audio data is sent to Meta’s servers for processing
• Ayumi shows a consent dialog before the first cloud transcription
• No data is stored on Ayumi’s servers — the request goes directly from your device to Meta
• See Meta’s terms for the Model API for how Meta handles API data