Groq Speech to Text

Groq Speech to Text uses GroqCloud’s Whisper models for fast file transcription. It is a separate provider from xAI Grok: Groq has its own API key, endpoint, models, limits, and consent. Ayumi writes a titled transcript without a prompt.

Getting a Groq API Key

1. Sign in to the GroqCloud Console
2. Create an API key
3. 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 Groq
3. Star Groq Whisper (Fast) to make it the default, or pick it for one recording

Groq profiles use the base Groq key unless a profile sets an Override API Key. Keys stay in Keychain and are not stored in queued transcription tasks.

Models

Model	Use it for
Whisper Large V3 Turbo	The default; faster transcription for everyday recordings
Whisper Large V3	The accuracy-oriented option when latency matters less

Ayumi stores these as groq/whisper-large-v3-turbo and groq/whisper-large-v3 so they can never be mistaken for xAI Grok model identifiers.

Options

Option	What it does
Language	Leave it on Automatic for detection, or send a language hint such as Japanese or English.
Segment Timestamps	Writes a [mm:ss–mm:ss] range before each Whisper segment.
Names and Terms	Biases recognition with names, jargon, and product names from typed lines, a Markdown file, or both.

Groq exposes Whisper’s context prompt rather than a dedicated vocabulary list. Ayumi merges file terms first, removes duplicates, keeps whole terms in order, and sends only what fits the prompt budget.

Uploads and Limits

• Ayumi sends supported recordings directly to Groq’s /openai/v1/audio/transcriptions endpoint.
• Direct uploads are limited to 25 MB; larger recordings are rejected before upload.
• Supported file extensions are FLAC, MP3, MP4, MPEG, MPGA, M4A, OGG, WAV, and WebM.
• Ayumi currently uses file transcription, not streaming, batch jobs, translation, or diarization.

See Groq Speech to Text documentation for current models, formats, and limits.

Data Privacy

When using a Groq profile:

• Audio is sent directly from your device to Groq for processing
• Ayumi asks for Groq-specific consent; consent previously given to Google, Meta, or xAI does not apply
• No audio passes through or is stored on Ayumi’s servers
• Groq says inference data is not retained by default, while limited reliability or abuse-monitoring data may be retained for up to 30 days unless Zero Data Retention applies; see Groq’s data documentation