Dictation API

The Dictation API turns a spoken clip into finished, formatted text in a single API call. It runs transcription on Universal-3.5 Pro and an LLM rewrite pass inside the same request, so what comes back is text your user could send without editing. Typical short clips return with a p50 of 134 ms.

A speech-to-text API returns a verbatim transcript, filler words and false starts included. The Dictation API returns the words your user meant, cleaned up and formatted to instructions you send with the request. Reach for the Sync Speech-to-Text API instead when you need the transcript itself, or when you need speaker diarization, which the Dictation API does not do.

Typical short dictation clips come back in under a second, with a p50 of 134 ms. That figure covers the transcription and the rewrite pass together, so the text is finished when it arrives rather than a draft you post-process.

The Dictation API is $0.62 per hour of audio, and that rate already includes the LLM rewrite, so there is no separate token bill to forecast. Usage is metered per second, which means a ten-second clip is billed as ten seconds. Volume discounts are available against committed usage.

The Dictation API supports 19 languages: English, Spanish, German, French, Portuguese, Italian, Turkish, Dutch, Swedish, Norwegian, Danish, Finnish, Hindi, Vietnamese, Arabic, Hebrew, Japanese, Urdu, and Chinese. Code-switching is native, so a speaker can move between them mid-sentence without you declaring a language up front.

The Dictation API runs on Universal-3.5 Pro, our flagship speech model, which is built for the short-form audio dictation produces. Send keyterms with the request, such as product names, drug names, or provider names, and they come back spelled correctly.

Yes. Send formatting instructions with a request and the Dictation API applies them to that response, whether that means stripping filler, applying your house style, returning a structured clinical note, or translating. Instructions are per request, so one integration can return different formats for different parts of your product.

AssemblyAI offers a HIPAA BAA on request for customers processing PHI through the Dictation API. Clinical dictation is a supported use case. Pass your formulary and provider names as keyterms so drug names, dosages, and ICD codes come back spelled right.

Yes. Sign up with a work email and you get free credits to test the Dictation API. No credit card, no sales call. Add a card when you outgrow the credits.

Reach out through the Contact sales form and a team member will be in touch within one business day. Sales can also quote volume pricing and set up a HIPAA BAA.