Transcribe audio

OpenAI speech to text

Hosted audio transcription API that turns uploaded recordings or live audio into text, with older models for speaker labels and timestamps.

At a glance

Provider: OpenAI
Category: Transcribe audio
Type: Hosted API
Pricing model: Per minute of audio; $0.0045 a minute for gpt-transcribe3
Free tier: Not published
Authentication: API key sent as a bearer token9
Connect via: REST API at /v1/audio/transcriptions, SDKs for JavaScript, Python, Go, Java, C# and Ruby, CLI1
Read from the provider's pages: 2026-11-04

Pricing model: Per minute of audio (gpt-transcribe); per token plus an estimated per-minute figure for older models
Free tier: Not published
Card required for free tier: Not published

Pricing in full →Tools

Official MCP server: No
Address or package: Not published

Tools in full →Authentication

Who creates the account: Not published
Where to get a key: Not stated on the pages read; the authentication reference says keys are secret and must stay server-side
Other requirements: Free tier requires an allowed geography

Authentication in full →Setup

REST API: https://api.openai.com/v1/audio/transcriptions
MCP server: None found
OpenAPI spec: Not published

Setup in full →Limits

File size: 25 MB per upload; split larger recordings
Diarization chunking: For audio over 30 seconds with gpt-4o-transcribe-diarize, set chunking_strategy
Known speaker references: Up to 4 reference clips of 2 to 10 seconds

Limits in full →Privacy

Data location: Regional storage and processing in the US, Europe and other regions for /v1/audio/transcriptions
Retention: No abuse monitoring retention and no application state for /v1/audio/transcriptions
Used for training: No, unless you opt in

Privacy in full →

Read from OpenAI’s own pages on 2026-11-04. Every number links to the page a value came from. For agents, the same facts as markdown: /tools/transcribe-audio/openai-transcription.md