Beta

Windows is currently a public beta. The Windows API is already useful for local tooling, but the surrounding product is still in beta.

HTTP API

TypeWhisper includes a local REST API for integration with external tools and scripts.

Note: The API is disabled by default. Enable it in Settings > API and configure the port (default: 9876). The API only accepts connections from localhost.

Check Status

# Check if the API is ready

curl http://localhost:9876/v1/status

{
  "status": "ready",
  "engine": "parakeet",
  "model": "nvidia_parakeet-tdt-0.6b-v2",
  "supports_streaming": false,
  "supports_translation": true
}

Transcribe Audio

# Send an audio file for transcription

curl -X POST http://localhost:9876/v1/transcribe \

-F "file=@recording.wav" \

-F "language=en"

{
  "text": "Hello, world!",
  "language": "en",
  "duration": 2.5,
  "processing_time": 0.8,
  "engine": "parakeet",
  "model": "nvidia_parakeet-tdt-0.6b-v2"
}

Optional Parameters

language - ISO 639-1 code (e.g., en, de). Omit for auto-detection.
task - transcribe (default) or translate.
target_language - ISO 639-1 code for translation target language (e.g., de, fr). Used with Canary or Marian translation.

List Models

# Get available models (local + cloud)

curl http://localhost:9876/v1/models

{
  "models": [
    {
      "id": "nvidia_parakeet-tdt-0.6b-v2",
      "engine": "parakeet",
      "downloaded": true,
      "active": true
    },
    {
      "id": "nvidia_canary-180m-flash",
      "engine": "canary",
      "downloaded": true,
      "active": false
    },
    {
      "id": "whisper-large-v3",
      "engine": "groq",
      "downloaded": true,
      "active": false
    },
    {
      "id": "gpt-4o-transcribe",
      "engine": "openai",
      "downloaded": true,
      "active": false
    }
  ]
}

Cloud models appear once you have configured an API key for the respective provider in Settings.

Error Responses

The API returns standard HTTP status codes with a JSON error body:

{
  "error": "No model loaded",
  "code": "MODEL_NOT_LOADED"
}

Common Error Codes

400 - Missing or invalid file field, unsupported audio format, or invalid parameter value.
503 - No model is loaded. Download and activate a model in Settings first.
500 - Internal error during transcription. Check the app logs for details.