# Transcription API: turn audio and video into text

> A hosted transcription service turns audio and video files up to 1 GB into text, with speaker labels, word timings, and no charge for failed jobs.

Oct 12, 2026. Topics: Developers, Workflows. https://www.generalinput.com/changelog/2026-10-12-transcribe-service

Workflows and apps can now turn recordings into text with a built-in transcription service: meeting recordings, voicemails, interviews, podcast episodes, or any audio or video file. Like every built-in service, it needs no account of your own and bills to your workspace.

### How it works

- **Start a job.** Send a link to the audio, or upload the file itself (audio or video, up to 1 GB), and you get back a job ID.
- **Get the transcript.** Check the job for its status, full text, detected language, and audio length, plus speaker turns when speaker labels are on and word-level timings when you ask for them.
- **Better accuracy on your vocabulary.** Pass key terms, such as product names or industry jargon, to bias recognition: up to 1,000 on the most accurate model and 200 on the other.
- **Two models.** Choose between two AssemblyAI models, with the provider kept behind a seam so more can be added.
- **Fair billing.** The price is locked when a job starts, each job is billed once, and a failed job is free.

Use it from code workflows, app handlers, and sandbox scripts, and pair it with the [decision service](/changelog/2026-10-12-decision-service) to route calls by what was said.