AssemblyAI
Speech-to-text, speech understanding and voice-agent APIs for developers
Overview
AssemblyAI provides speech-to-text APIs for recorded and live audio, priced per hour, with add-ons for diarization, PII redaction, summaries, sentiment and topic detection. It also sells a Dictation API, a bundled Voice Agent API and an LLM Gateway for running language models over transcripts. It is an API for developers, not an end-user transcription app.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 14 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Keyterms prompting included
- Speaker diarization +$0.02/hr
- Native code switching
- Prompting and keyterms +$0.05/hr each
- Medical mode +$0.15/hr
- Keyterms prompting +$0.04/hr
- Speaker diarization +$0.12/hr
- Keyterms prompting included
- Voice Focus +$0.10/hr
- Realtime STT, Voice Agent LLM and TTS
- Turn and interruption detection
- Hosting and bring-your-own Twilio SIP
- Custom rate limits and enhanced concurrency
Use cases
What you can produce with AssemblyAI
- Transcripts of recorded audio or video with speaker labels, punctuation and word-level timestamps
- Live transcripts streamed from realtime audio for captions, notes or voice agents
- Summaries, sentiment, entities, topics and named speaker identification from speech-understanding add-ons
- Transcripts and audio with PII redacted
- Ready-to-send text from a spoken utterance through the Dictation API
- A hosted phone or web voice agent from the Voice Agent API
- Language-model answers and extraction over transcripts through the LLM Gateway
AssemblyAI is one of the more complete speech APIs: accurate recorded and live transcription, a long list of per-hour add-ons, clear published prices from $0.15 an hour, good documentation and a free starting allowance that is enough for a real evaluation. The newer Sync, Dictation and Voice Agent APIs make it a broader voice platform than a plain transcription service.
Model your full bill, because add-ons and multichannel audio stack on the base rate, and treat accuracy claims as vendor benchmarks. ASEAN-language accuracy needs your own testing, especially on the Pro models, which cover fewer languages. Developers only.
What we found
AssemblyAI sells speech AI to developers as an API rather than as an app. Dylan Fox started the company in late 2017, and its last disclosed raise was a $50 million Series C led by Accel in December 2023, with Insight Partners and Y Combinator taking part, which TechCrunch reported brought total funding to $115 million. No later round has been announced. It works remote-first, and the operating entity is AssemblyAI Inc.
The model line-up has moved on from what earlier versions of this page described. For recorded audio, Universal-3.5 Pro is the flagship at $0.21 an hour, covering 18 languages with code switching, and Universal-2 stays at $0.15 an hour across 99 languages. For live audio, Universal-3.6 Pro Realtime superseded Universal-3.5 Pro Realtime on 29 September 2026, at $0.45 an hour in 32 languages, while English-only Universal-Streaming and a six-language multilingual version cost $0.15. There is also a Sync API for clips up to two minutes, a Dictation API, and a Voice Agent API at $4.50 an hour that bundles recognition, an in-house language model, text-to-speech and hosting. LeMUR, the old way to run language models over transcripts, now redirects to an LLM Gateway that resells about fifty third-party models at token prices.
The headline rate is rarely the whole bill. Speaker diarization, keyterm prompting, medical mode, PII redaction and the speech-understanding models for summaries, sentiment, entities and topics are each priced per hour on top, and multichannel audio is billed per channel, so an hour of stereo counts as two. New accounts get a free allowance of up to 185 hours of recorded or 333 hours of streaming transcription with no card, after which pay-as-you-go is invoiced monthly with no minimum.
Accuracy claims on the site come from AssemblyAI's own benchmarks, so read them as vendor figures. ASEAN teams should test Bahasa, Thai, Vietnamese or Tagalog audio themselves, and note that the newer Pro models list far fewer languages than Universal-2.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including AssemblyAI's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with AssemblyAI unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to AssemblyAI directly →
Spotted something out of date? Suggest an update →
Alternatives to AssemblyAI
More in Video & Audio
