Jobs (Async Transcription)

Audio files longer than 3 minutes are processed asynchronously. When you submit long audio to the transcription endpoint, you receive a job ID instead of an immediate result. Poll the job status endpoint to retrieve your transcript when processing completes.

How It Works

Submit audio via POST /transcribe as usual.
If the audio exceeds 3 minutes, the response returns a job_id with status queued.
Poll GET /jobs/{job_id} until status is completed or failed.

Submission Response

When audio is queued for async processing, the transcribe endpoint returns:

{
  "job_id": "job_a1b2c3d4e5",
  "status": "queued",
  "audio_duration_seconds": 245.6,
  "message": "Audio exceeds 3 minutes. Processing asynchronously. Poll GET /jobs/{job_id} for results."
}

Poll Job Status

GET /jobs/{job_id}

Requires the same API token used to submit the audio. Returns the current job status.

Job Statuses

Status	Description
queued	Job is waiting to be picked up by a worker
processing	Audio is currently being transcribed
completed	Transcription finished, result available
failed	Processing failed, error details available

Example

# 1. Submit long audio (returns job_id)
curl -X POST "https://api.ayaspeech.ayadata.ai/transcribe" \
  -H "x-api-token: your_api_token" \
  -F "audio_file=@long_recording.wav"

# 2. Poll for results
curl "https://api.ayaspeech.ayadata.ai/jobs/job_a1b2c3d4e5" \
  -H "x-api-token: your_api_token"

Completed Response

{
  "job_id": "job_a1b2c3d4e5",
  "status": "completed",
  "result": {
    "text": "Medaase paa, me din de Kwame. Me firi Kumasi...",
    "language": "twi",
    "inference_time_ms": 8542.3
  }
}

Failed Response

{
  "job_id": "job_a1b2c3d4e5",
  "status": "failed",
  "error": "Worker timeout: transcription exceeded maximum processing time."
}

Notes

The async threshold is 3 minutes of audio. Shorter files are transcribed synchronously.
Jobs expire after 24 hours, poll results before then.
Only the API token that submitted the job can poll its status.
Recommended polling interval: every 5–10 seconds.
Usage (minutes) is deducted at submission time, not when the job completes.