AssemblyAI

Speech-to-text API for pre-recorded and streaming audio, with speaker labels, language detection, speech understanding and voice agents. SDKs for Python and JavaScript/TypeScript.

Website Docs Changelog Compare with others

AssemblyAI at a glance

From official pages, checked Nov 15, 2026. Tap the arrow for each source.

Free tier: Yes
$50 in free credits at signup, no credit card required.
Starting price: $0.15 per hour
Universal-2 pre-recorded speech-to-text (Universal-3.5 Pro: $0.21/hr).
Rate limit: 5 concurrent transcriptions
Free accounts, async; paid accounts get 200+.
Size limit: 5 GB / 10 hours per file
Transcript endpoint; direct uploads via /v2/upload capped at 2.2 GB.
Storage: Transcripts kept ~30 days
Async audio deleted within 24-48 hours; transcript deletion begins at 30 days.
Official SDKs: PythonJavaScript / TypeScript
Trains on your data: Opt-out
Certain PII-redacted files used for training unless opted out, on a BAA, or using EU servers.
EU hosting: Yes
EU endpoint api.eu.assemblyai.com keeps data in the EU.
Compliance: SOC 2ISO 27001GDPRHIPAA
SOC 2 Type 2 and ISO 27001; GDPR via EU servers; HIPAA BAA available.

What you can build with AssemblyAI

AssemblyAIPro archiveSnippetPaste

Transcribe audio with speaker labels using AssemblyAI in Python

Get a transcript split by speaker, with automatic language detection, from any audio URL using the AssemblyAI Python SDK.

Free tier: YesFrom $0.15 per hour

SummariesVoice

Recent updatesAll updates

Nov 13, 2026
AssemblyAI LLM Gateway adds a decisions endpoint for yes/no, choice and score answers; Python SDK 1.10.0 supports it (November 12, 2026)
AssemblyAI's LLM Gateway has POST /v1/decisions: typed answers with probabilities. Python SDK 1.10.0 adds decisions.create(). US only.
Nov 11, 2026
AssemblyAI Vet Mode beta: set domain to veterinary-v1 for vet transcripts (November 10, 2026)
AssemblyAI added Vet Mode in beta. Set domain to veterinary-v1 on Universal-3.5 Pro, Universal-2 or the streaming models.
Nov 7, 2026
AssemblyAI launches an MCP server to transcribe and analyze audio from Claude, Cursor and VS Code (November 3, 2026)
Add mcp.assemblyai.com/mcp to your MCP client: five tools to submit, read, summarize and analyze transcripts at normal API rates.
Oct 28, 2026
AssemblyAI launches AssemblyAI-TTS-1 text-to-speech for voice agents at $15 per 1M characters (October 27, 2026)
AssemblyAI's first TTS model is generally available: streaming audio, 4,000+ voices, six languages, $15 per 1M characters.