Skip to main content

Transcription

When you save a Listen recording, Toby can transcribe audio with a configured transcription provider. These settings are separate from your chat AI provider (though OpenAI, Vercel AI Gateway, and OpenRouter can reuse the same API keys you set under AI).

Open Toby.app → Settings → Transcription.

Settings​

SettingPurpose
ProviderTranscription backend (OpenAI, Groq, Vercel AI Gateway, or OpenRouter)
ModelModel id offered by that provider
API KeyOptional dedicated key for transcription
Persona for recording summariesPersona used when you summarize a recording transcript (falls back to the default persona)

OpenAI​

If the provider reuses OpenAI, you can leave the transcription API key empty and Toby will use Settings → AI → OpenAI → API Token when no transcription-specific key is set. The Settings UI shows a hint when that applies.

Vercel AI Gateway​

The Vercel AI Gateway provider routes transcription requests through the same gateway key you use for chat. It reuses Settings → AI → Vercel AI Gateway → API Key (or AI_GATEWAY_API_KEY / VERCEL_OIDC_TOKEN) when no transcription-specific key is set.

Available models are fetched from the live gateway catalog (filtered to type: transcription). The curated fallback list includes:

SlugUpstream
openai/whisper-1OpenAI Whisper
openai/gpt-4o-mini-transcribeOpenAI GPT-4o mini transcribe
openai/gpt-4o-transcribeOpenAI GPT-4o transcribe
xai/grok-sttxAI Grok STT

Note: Gateway transcription is non-streaming and subject to the same audio payload limits as other providers (25 MB before chunking).

OpenRouter​

The OpenRouter provider uses OpenRouter’s speech-to-text API (/api/v1/audio/transcriptions), which is OpenAI-compatible. It reuses Settings → AI → OpenRouter → API Key (or OPENROUTER_API_KEY) when no transcription-specific key is set.

Available models are loaded from the live OpenRouter catalog:

GET https://openrouter.ai/api/v1/models?output_modalities=transcription

If the catalog is unreachable, Toby falls back to a curated list (Whisper, GPT-4o transcribe variants, Voxtral Mini Transcribe, Deepgram Nova-3, and similar).

Note: Multipart uploads are limited to 25 MB (same as other providers); Toby chunks larger recordings automatically when possible.

When transcription runs​

  • After you stop and save a recording in Toby.app (or a capture path that finalizes into Listen storage)
  • When you retry / re-transcribe a recording from the Recordings window or related tools
  • Via the local service Server API listen transcribe endpoint used by the app

Capture itself uses Toby.app’s Native API (microphone / system audio). Listing and transcription use the daemon Server API.

Recording summaries​

Once a recording has a transcript, Summarize / Re-Summarize in the Recordings inspector generates an AI summary using the persona selected here. See Listen mode.

If transcription fails​

  1. Confirm Provider, Model, and key (or OpenAI AI token).
  2. Open Recordings, select the recording, and retry transcription.
  3. Check that the recording has playable audio (combined.m4a or source WAVs).