Turn recorded or live audio into accurate text through a pay-as-you-go speech-to-text API.
About AssemblyAI
AssemblyAI is a speech-to-text API that transcribes recorded and live audio for developers. It suits teams building voice agents, meeting tools and call analytics, and it starts free with no credit card.
AssemblyAI is a speech-to-text API for developers. Its Universal-3.5 Pro model transcribes recorded audio in 18 languages, while Universal-2 covers 99 languages. A realtime model handles live audio for voice agents, a Sync API returns short clips in one call, and add-ons cover speaker labels, keyterm prompting and a medical mode. It also sells a streaming text-to-speech model.
It is built for developers and product teams adding voice features to apps, not for people who just want to upload a file and read a transcript. Costs can add up, since add-ons are billed per hour on top of the base rate. The free tier needs no card, and paid use is available on a pay-as-you-go basis.
- Category
- ai audio voice, AI Productivity
- Pricing model
- Freemium, Paid, Contact Sales
- Starting price
- Free tier; pay as you go from $0.15/audio hour
- Platforms
- API / SDKWeb
- API
- Available
- Made for
- App developersVoice agent buildersMeeting and call analytics teamsHealthcare software teams
What you get
Universal-3.5 Pro transcribes recorded audio in 18 languages and handles people switching languages.
Universal-2 supports 99 languages for $0.15 per hour of audio.
The realtime model transcribes live audio in 32 languages for voice agents, at $0.45 an hour.
Send a clip up to 20 minutes to the Sync API and get the full transcript back in one call.
Speaker diarization splits a transcript by who said what, for $0.02 an hour.
Add up to 1,000 keyterms, like product names, so the model spells them right.
The good and the bad
What's good4
- The free tier needs no credit card and covers up to 185 hours of recorded audio.
- Pay-as-you-go pricing has no monthly commitment, so small projects stay cheap.
- A medical mode improves accuracy on healthcare terms for $0.15 an hour extra.
- One account also covers dictation and streaming text-to-speech, not only transcripts.
What's not3
- It is an API, so non-developers need another app on top to use it.
- Add-ons like prompting and diarization are billed separately, which makes costs harder to predict.
- The most accurate recorded-audio model covers 18 languages, far fewer than Universal-2's 99.
Pricing
- Realtime $0.45/hr
- Sync API $0.30/hr
- Add-ons billed per hour
User Reviews & Ratings
Reviews from people who use AssemblyAI. Sign in with Google to add yours.
Questions people ask
What is AssemblyAI and who is it for?
AssemblyAI is a speech-to-text API that turns audio into text. It is for developers building products like voice agents, note takers and call analytics.
Is AssemblyAI free?
Yes, you can sign up and start without a credit card. The free tier includes up to 185 hours of recorded transcription and up to 333 hours of streaming transcription.
How much does AssemblyAI cost?
Recorded audio costs $0.21 an hour with Universal-3.5 Pro or $0.15 with Universal-2. Realtime transcription is $0.45 an hour, the Sync API is $0.30, and dictation is $0.62. Custom plans are available from sales.
What add-ons cost extra?
On Universal-3.5 Pro, keyterm prompting and plain-language prompting cost $0.05 an hour each. Speaker diarization is $0.02 an hour, and medical mode is $0.15 an hour.
How many languages does AssemblyAI support?
Universal-2 supports 99 languages. Universal-3.5 Pro covers 18 languages for recorded audio, and the realtime model covers 32.
Can it transcribe live audio?
Yes. The Realtime Speech-to-Text API streams transcripts at low latency, with speaker labels and keyterm prompting, for $0.45 an hour.
Does AssemblyAI offer text-to-speech?
Yes. AssemblyAI-TTS-1 streams speech with about 90 ms to first audio in English, Spanish, French, German, Italian and Portuguese, priced at $15 per million characters.
What are AssemblyAI's main limitations?
You need some coding to use it. Its top recorded-audio model supports fewer languages than its older model, and add-ons raise the per-hour price.
How does AssemblyAI compare to OpenAI Whisper?
Whisper is a model you can run yourself, while AssemblyAI is a hosted API with speaker labels, realtime streaming and add-ons built in. AssemblyAI charges per audio hour.
Is AssemblyAI easy for beginners?
It is easy for developers, since a short API call returns a transcript. People who do not code will find it hard to use on its own.
Alternatives to AssemblyAI
See allTurn text into lifelike speech, clone voices and dub audio into other languages with ElevenLabs.
Free plan, paid from $6/month
Remove background noise from calls and get AI meeting notes, transcripts and action items.
Free plan, paid from $16/month
Capture every sales call and email, spot deal risks and automate follow-ups with Gong revenue AI.
No free plan, price on request
Connect thousands of apps and automate work with Zaps, AI agents, Tables and Forms, no code needed.
Free plan, paid from $19.99/month
