Transcription API

Quick start

curl -X POST https://getsrt.app/api/v1/transcripts \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -F "audio=@/path/to/podcast.mp3"

{
    "job_id": "9b1deb4d-3b7d-4bad-9bdd-2b0d7b3dcb6d",
    "status": "pending",
    "check_status_url": "https://getsrt.app/api/v1/transcripts/9b1deb4d-..."
  }

curl -L https://getsrt.app/api/v1/transcripts/9b1deb4d-... \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -o transcript.srt

{ "status": "processing", "progress": 66, "completed_chunks": 2, "total_chunks": 3 }

Good to know

• A valid API token is required for every request — anonymous uploads are rejected with 401.
• One credit is charged per second of audio, only on a successful transcription.
• You must have enough credits to cover the audio duration, or the upload is rejected with 402.
• Re-running a failed job is always free.
• Long files are split at natural silences and transcribed in parallel for speed.
• Audio defaults to English; pass source_language to transcribe another language or auto-detect it (see below).

Poll until done

job_id=9b1deb4d-...
until curl -sf https://getsrt.app/api/v1/transcripts/$job_id \
        -H "Authorization: Bearer YOUR_API_TOKEN" -o transcript.srt \
        && head -c3 transcript.srt | grep -q "1"; do
    echo "still working…"; sleep 1;
done
echo "saved transcript.srt"

Specify a source language

By default the audio is assumed to be English. Add two optional fields to the same submit request to transcribe another language, or to translate the speech to English:

• source_language — either auto (let whisper detect it) or one of the language codes listed below. Defaults to en.
• translate_to_english — 1 to translate the source speech to English (whisper's only supported translation target). Omit to keep the transcript in the source language.

curl -X POST https://getsrt.app/api/v1/transcripts \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -F "audio=@/path/to/podcast.mp3" \
  -F "source_language=zh" \
  -F "translate_to_english=1"

The downloaded .srt filename carries a <source>-<target> suffix so you can tell jobs apart at a glance — e.g. podcast.zh-en.srt for Chinese audio translated to English, or podcast.en-en.srt when no translation was requested.