Speech-to-text API
AssemblyAI Speech to Text
Speech to text over a REST API for recorded audio and a WebSocket for live audio, billed per hour of audio or of connection time.
Plans
Values that apply to one plan. "Not stated" means the official pages do not say; it is not an estimate.
| Plan | Price | Included | Extra usage | Concurrency | Rate limit | Buy |
|---|---|---|---|---|---|---|
| Pay as you go, pre-recorded (Universal-2)Pay as you go | Not recorded | Not recorded |
Not recorded |
Pricing pageChoose the plan on the page | ||
| Pay as you go, streaming (Universal-Streaming)Pay as you go | Not recorded | Not recorded |
Not recorded |
Pricing pageChoose the plan on the page |
Applies to every plan
Recorded for the product or the provider; a plan-level value or a listed exception replaces it.
Collection notes
Pay as you go only. The pre-recorded plan is the Universal-2 model at $0.15 an audio hour (Universal-3.5 Pro is $0.21) and the streaming plan is Universal-Streaming at $0.15 an hour (Universal-3.6 Pro Realtime is $0.45). The EU region costs the same as the US region. Streaming is billed on the time the connection stays open, not on the audio sent. The free tier is a one-time $50 credit at sign-up (about 185 hours pre-recorded or 333 hours streaming). The docs MCP server is for AI coding assistants and does not call the transcription API.
Sources and notes
Each number points to the provider's own sentence, as it appears on the page on the date shown.
EU region is the same price as US. Use `api.eu.assemblyai.com` / `streaming.eu.assemblyai.com` for GDPR-compliant data residency at no premium.
The EU endpoint (`api.eu.assemblyai.com`) guarantees your data never leaves the European Union.
The pricing page names the US and the EU regions and prices them the same.
Pre-recorded transcription is billed on the duration of the submitted audio or video file in seconds, multiplied by the hourly rate for the selected speech model.
This price line is for pre-recorded audio; live audio has its own price line (the streaming plan).
| Diarization, async — standard | +$0.02/hr | Universal-3.5 Pro async, Universal-2 async, SLAM-1 |
Speaker diarization is an add-on of $0.02 an audio hour on pre-recorded audio.
How the value follows from the source: $0.02 per audio hour is $20 per 1,000 audio hours.
Add-on features (for example, Entity Detection or Medical Mode) are billed separately at their own hourly rates, also pro-rated to the exact second.
Billing is by the seconds of the file; no sentence names a minimum duration.
Hindi, Croatian, Hungarian, Korean, Macedonian
Korean is in the supported languages of pre-recorded transcription, in the group with the higher word error rates.
| Universal-2 | `universal-2` | **$0.15/hr** |
How the value follows from the source: $0.15 per audio hour is $150 per 1,000 audio hours.
How the value follows from the source: $0.15 per audio hour.
Streaming Speech-to-Text is billed on the total duration that your WebSocket connection stays open
This price line is for live audio; recorded audio has its own price line (the pre-recorded plan).
| Diarization, streaming | +$0.12/hr | All streaming models incl.
Speaker diarization is an add-on of $0.12 an hour on streaming.
How the value follows from the source: $0.12 per hour is $120 per 1,000 hours.
Streaming Speech-to-Text is billed on the total duration that your WebSocket connection stays open, not on the amount of audio you send.
Streaming is billed on the time the connection stays open, which is not the audio time and is not stated as a per-request minimum or the absence of one.
| Universal-Streaming English | `universal-streaming-english` | **$0.15/hr** |
Universal-Streaming is described as an English-only model; the pages read do not state Korean for any streaming model.
How the value follows from the source: $0.15 per hour is $150 per 1,000 hours.
How the value follows from the source: $0.15 per hour of connection time.