What is AssemblyAI?
AssemblyAI is a speech AI platform sold as APIs for pre-recorded and real-time speech-to-text, a Voice Agent API, speech understanding add-ons and an LLM Gateway. Developers call it over web APIs (including WebSocket for streaming); it is not a consumer app.
Who it's for
It suits developers and companies that build transcription, call analysis, meeting or voice-agent features into their own products. The vendor sells it as pay-as-you-go APIs with enterprise contracts for higher concurrency, so it is not aimed at people who want a ready-made transcription app.
Pros & cons
Strengths
- The pricing page lists pre-recorded speech-to-text at $0.21 an hour for Universal-3.5 Pro and $0.15 an hour for Universal-2, with no minimums, upfront fees or contracts on pay as you go.
- The models reference lists 99+ languages for Universal-2 and 18 for Universal-3.5 Pro on pre-recorded audio, and 32 languages for Universal-3.6 Pro Realtime on streaming.
- The Voice Agent API runs speech-to-text, an LLM and text-to-speech over one WebSocket, and the pricing page lists it at $4.50 an hour ($0.075 a minute), billed per second of connected time, with the LLM, TTS, turn detection, hosting and SIP trunking included.
- Speech understanding add-ons are priced per hour, for example Translation at $0.06, Summarization at $0.02 (low effort), Entity Detection at $0.08 and PII Text Redaction at $0.08; the vendor says Translation covers 100+ target languages.
- The vendor lists SOC 2 Type 2, ISO 27001, GDPR, PCI-DSS and a HIPAA BAA, plus EU endpoints at the same price, and states that these carry no premium pricing.
- The LLM Gateway offers a single endpoint to 25+ models from five providers, with automatic provider fallback, structured JSON outputs and prompt caching, per the models reference.
Limitations
- Billing is usage-based, billed monthly in arrears, and add-ons stack on top of the base rate (for example diarization, prompting or Medical Mode), so the final hourly cost depends on which features are switched on. Multichannel audio is billed per channel, so a 1-hour stereo file counts as 2 hours.
- The vendor's pages disagree on the free tier: the pricing page says the Free plan includes 185 hours of pre-recorded and 333 hours of streaming transcription, while the billing docs and the machine-readable pricing page describe $50 in free credits with no hour figures. Confirm which applies at signup.
- Free streaming is limited to 5 new connections per minute; pay as you go starts at 100 sessions per minute and rises 10% automatically when usage reaches 70% of the limit. Higher starting limits are handled through sales.
- The billing docs say LLM Gateway is not covered by free credits, and that API access pauses when the balance reaches $0 without auto-pay. Pricing for LLM Gateway is 10% higher for in-region US or EU routing than for global routing.
- Data Controls (retention TTL and the model-training opt-out) requires a paid plan, and the default training setting is not stated on the page read. Single sign-on is a paid add-on at $199 per connection per month, and the pages read do not describe a refund policy.
Plans and pricing
- Free: $0. No credit card required. The pricing page lists 185 hours pre-recorded and 333 hours streaming; the billing docs describe $50 in free credits that do not expire. Streaming limited to 5 new connections per minute. LLM Gateway not covered by free credits.
- Pay as you go: From $0.15 an hour (usage-based). Pre-recorded: Universal-2 $0.15 an hour, Universal-3.5 Pro $0.21 an hour. Streaming: Universal-Streaming $0.15 an hour, Universal-3.6 Pro Realtime $0.45 an hour. Voice Agent API $4.50 an hour. Add-ons and LLM Gateway token rates extra. No minimums or contracts; starts at 100 streaming sessions per minute.
- Enterprise / custom: Custom pricing. Custom rate limits, higher concurrency and volume pricing through sales. Specific terms are not published on the pricing page.
Prices were read from the official pricing page, the machine-readable pricing page (last updated 2026-09-28) and the billing docs, all in US dollars and billed monthly in arrears. The free-tier allowance differs between the pages (185/333 hours versus $50 in credits). Confirm on the live pricing page.
Details checked October 2026 against the vendor's own pages. Prices and limits change, so confirm before you buy. Official pricing page ↗
Sources
This profile was compiled from these public pages. We have not run hands-on tests.
AssemblyAI — at a glance
AssemblyAI is a speech AI platform sold as APIs for pre-recorded and real-time speech-to-text, a Voice Agent API, speech understanding add-ons and an LLM Gateway. Developers call it over web APIs (including WebSocket for streaming); it is not a consumer app.
Try AssemblyAI →