Synchronous transcription

curl https://api.omi.health/v1/audio/transcriptions \
  -H "Authorization: Bearer $OMI_API_KEY" \
  -F file=@dictation.m4a \
  -F model=omi-medical-1 \
  -F language=auto \
  -F response_format=verbose_json

Authenticate every request with Authorization: Bearer <key>. The file may be WAV, MP3, M4A/MP4, WebM, OGG, FLAC, or AAC; Omi converts it to mono 16 kHz PCM internally. Direct uploads inspect the container when a client sends a common MIME alias such as audio/x-m4a,audio/m4a, or audio/mp4a-latm.

curl https://api.omi.health/v1/audio/transcriptions \
  -H "Authorization: Bearer $OMI_API_KEY" \
  -F file=@dictation.m4a \
  -F model=omi-medical-edge-1 \
  -F language=en \
  -F diarize=true \
  -F max_speakers=2 \
  -F response_format=verbose_json

Hosted Edge returns deterministic transcript segments and word timestamps for both short and long audio, with optional speaker labels when diarize=true. It does not support realtime, vocabulary, Patterns, language detection, or processing profiles. Excess Edge demand may queue or return a retryable capacity response so it cannot slow flagship traffic.

curl https://api.omi.health/v1/audio/transcriptions \
  -H "Authorization: Bearer $OMI_API_KEY" \
  -F file=@consultation.wav \
  -F model=omi-medical-1 \
  -F language=en-AU \
  -F response_format=verbose_json \
  -F diarize=true \
  --form-string 'vocabulary=["Tinel","Hepcludex","tirzepatide"]'

For best vocabulary behavior, provide 5–50 exact terms likely to be spoken in this recording. Lists of 50 terms or fewer are considered in full; larger lists select the 50 most relevant terms. Speaker labels and timestamps never change the transcript wording.

Eligible ordinary requests can return synchronously up to 15 minutes without enabling zero retention. The usual 60-second audio boundary applies to requests outside that extended inline path; it is not an HTTP timeout. Both limits measure recording duration, not upload or processing time.

For extended inline delivery, send omi-medical-1, language=en, dictionary=falseand a json, text or verbose_json response. Keep speakers, custom vocabulary, context, webhooks and idempotency keys off. Audio must fit the 32-MiB RAM budget, including its decoded analysis audio. General, speakers and other excluded options on long recordings use async delivery. Ordinary integrations should handle both 200 and 202.

curl https://api.omi.health/v1/audio/transcriptions \
  -H "Authorization: Bearer $OMI_API_KEY" \
  -F 'file=@dictation.flac;type=audio/flac' \
  -F model=omi-medical-1 \
  -F language=en \
  -F dictionary=false \
  -F response_format=json

Add X-Omi-Retention: none to require RAM-only processing and prohibit stored fallback. The header does not unlock the 15-minute sync limit. For connection timeouts, see the production guide.

Available on Builder, PAYG and Enterprise at the same transcription price. Send X-Omi-Retention: none to process in memory and receive the transcript on this connection, without storing audio or a result for later retrieval. The console setting applies this requirement to all your batch keys; it can take up to 60 seconds to propagate, so use the header when you need it immediately.

curl https://api.omi.health/v1/audio/transcriptions \
  -H "Authorization: Bearer $OMI_API_KEY" \
  -H "X-Omi-Retention: none" \
  -F file=@dictation.m4a \
  -F model=omi-medical-1 \
  -F language=en \
  -F dictionary=false \
  -F response_format=json

This mode supports English omi-medical-1, explicit language=en and dictionary=false, with json, text or verbose_json responses, up to 15 minutes and the 32-MiB memory budget. Speakers, custom vocabulary, context, webhooks and idempotency keys are unsupported. Unsupported requests return 422; busy capacity returns 503. Neither falls back to stored jobs. A successful response acknowledges X-Omi-Retention: none.

If the connection fails, there is no saved result to fetch again. Content-free usage and audit metadata remain. For durable async delivery, choose your retention window or delete the result using DELETE /v1/jobs/{job_id}/result.