Cloud STT API

Base path: /api/stt

Cloud speech-to-text transcription using Groq Whisper or OpenAI Whisper. The API key stays server-side — the browser never contacts the STT provider directly.

---

POST /api/stt/transcribe

Transcribe an audio recording. Accepts a multipart upload and forwards it to the configured cloud STT provider.

Request

POST /api/stt/transcribe
Content-Type: multipart/form-data

Field	Type	Description
audio	File	Audio recording (webm, ogg, wav, mp3, etc.)

Constraints:

• Maximum file size: 25 MB (Groq/OpenAI hard limit)
• Minimum size: 100 bytes (rejects empty/noise-only recordings)

Response

{
  "ok": true,
  "data": {
    "text": "The transcribed text from the audio recording."
  }
}

Error Responses

Status	Error
400	Cloud STT not configured — set provider in Settings → Voice
400	No API key configured for the provider
400	No audio file provided
400	File too large (> 25 MB)
400	Audio too short (< 100 bytes)
502	Network error calling STT provider
502	STT provider returned an error

Example (curl)

curl -X POST http://localhost:3000/api/stt/transcribe \
  -F "audio=@recording.webm"

---

Configuration

Configure the cloud STT provider in ~/.getthatquick/config/settings.json (or via Settings → Voice in the UI):

{
  "stt": {
    "provider": "groq",
    "cloudApiKey": "gsk_...",
    "cloudModel": "whisper-large-v3-turbo"
  }
}

Field	Values	Description
provider	"local", "groq", "openai-whisper"	STT engine to use
cloudApiKey	string	API key for the cloud provider
cloudModel	string	Model identifier (see below)

Available Models

Groq (recommended — free tier):

Model	Speed	Accuracy
whisper-large-v3-turbo	Fastest	Good
whisper-large-v3	Slower	Best

OpenAI Whisper:

Model	Notes
whisper-1	Classic model
gpt-4o-transcribe	Best accuracy
gpt-4o-mini-transcribe	Faster, cheaper