Speech-to-text and speech understanding APIs priced per audio hour.
What it does
On this page
AssemblyAI sells speech models behind an API. The range covers recorded and realtime transcription, speaker diarization, summarisation, sentiment, entity detection, PII redaction and a newer text-to-speech model.
How it works
You send audio or a stream to an endpoint and get structured JSON back, with word timings and speaker labels where the feature is enabled. The same platform offers a sync endpoint for short requests and a dictation endpoint for single utterances.
Who it is for
Developers building notetakers, call analytics, voice agents or medical and legal transcription, and teams that need the model inside their own network.
Pricing
Pay as you go, with a free start. Rates are per audio hour by model: $0.21 for the flagship recorded model, $0.30 for realtime, and $0.62 for dictation. Text-to-speech is $15 per million characters.
Limits
Each extra feature is priced on top of the base transcription rate, so a full pipeline costs more than the headline figure. Self-hosting is a separate deployment agreement.
Source: https://www.assemblyai.com/