AssemblyAI launches single-call Dictation API at $0.62/hr

Dictation apps usually do two jobs: transcribe the speech, then clean it up. In the demo video for Blurt, its push-to-talk dictation app, AssemblyAI says these are typically built as two steps: a speech-to-text model, then a separate LLM that strips out the ums and tidies the text. Its Dictation API does both in one request, and the instructions for how the result should read go in that same request. The app, the presenter says, is built on that one endpoint.

The post describes the output: a spoken clip comes back as finished text with filler words and false starts removed, formatted as notes, a commit message or a reply to a customer. In the video, the presenter dictates into Google Docs, switching between English, Spanish and Hindi mid-recording, and fills a key-terms setting with unusual names and drug names that he says then come out spelled correctly. That is a single take, so it says nothing about how reliably it works.

AssemblyAI says the API runs on its Universal-3.5 Pro model, supports 19 languages, handles short clips in under a second and costs $0.62 per hour of audio. These are the company's own figures; the video's narration says 18 languages, not 19. AssemblyAI's launch post a day earlier showed the same API filling a multi-field form from one spoken answer.