# 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.

- Provider: OpenAI (https://openai.com)
- Category: Transcribe audio
- Type: Hosted API
- Pricing model: Per minute of audio; $0.0045 a minute for `gpt-transcribe` [3]
- Free tier: Not published
- Authentication: API key sent as a bearer token [9]
- Connect via: REST API at `/v1/audio/transcriptions`, SDKs for JavaScript, Python, Go, Java, C# and Ruby, CLI [1]
- Read from the provider's pages on 2026-11-04
- Page: https://supermarket.dev/tools/transcribe-audio/openai-transcription

## Pricing

### Billing
- Pricing model: Per minute of audio (`gpt-transcribe`); per token plus an estimated per-minute figure for older models [4]
- Free tier: Not published
- Card required for free tier: Not published
- Minimum spend: Build tier starts at $5 in total credit purchases [7]
- Billing period: Monthly usage limit per tier, for example $500 a month at Build [7]

### Prices
Prices are as listed on the pricing page. Per-token columns are dollars per 1M tokens (input, output); the last column is the page's estimated cost.
| Model | Type | Input / 1M tokens | Output / 1M tokens | Per hour of audio |
|---|---|---|---|---|
| `gpt-transcribe` (recommended) | Transcription | - | - | $0.27 (= $0.0045 x 60) |
| `gpt-4o-transcribe` (deprecated) | Transcription | $2.50 | $10.00 | $0.36 (= $0.006 x 60) |
| `gpt-4o-mini-transcribe` (deprecated) | Transcription | $1.25 | $5.00 | $0.18 (= $0.003 x 60) |
| `gpt-4o-transcribe-diarize` (deprecated) | Transcription + diarization | $2.50 | $10.00 | $0.36 (= $0.006 x 60) |
| `whisper-1` (deprecated, listed as Whisper) | Transcription | - | - | $0.36 (= $0.006 x 60) |
| `gpt-live-transcribe` (live audio) | Live transcription | - | - | $1.02 (= $0.017 x 60) |
Source [4]

## Tools

### MCP server
- Official MCP server: No [13]
- Address or package: Not published

### API
| Endpoint | What it does |
|---|---|
| `POST /v1/audio/transcriptions` | Transcribes an uploaded file; `stream=true` streams partial text |
| `/v1/audio/translations` | Translates a recording into English with `whisper-1`; deprecated |
| `/v1/realtime/transcription_sessions` | Transcribes live audio over a Realtime session |
Source [3]

### Features
- Default model: `gpt-transcribe` [1]
- Languages: `gpt-transcribe` takes expected languages as ISO 639-1 codes, selected ISO 639-3 codes and regional `zh` codes; `whisper-1` supports 98 languages [1]
- Speaker labels: Not on `gpt-transcribe`. Only the deprecated `gpt-4o-transcribe-diarize` returns them, using `response_format` `diarized_json`; shutdown is Feb 26, 2027 [2]
- Word timestamps: Not on `gpt-transcribe`. Only the deprecated `whisper-1` supports `timestamp_granularities[]`, plus `srt` and `vtt` subtitles [1]
- Live streaming: File transcription streams partial text with `stream=true` on `gpt-transcribe`; live audio uses `gpt-live-transcribe` over Realtime, which returns no word timestamps or speaker labels [1]
- Max file size: 25 MB [1]
- Audio input: `mp3`, `mp4`, `mpeg`, `mpga`, `m4a`, `wav`, `webm` [1]
- Deprecated models: `whisper-1`, `gpt-4o-transcribe`, `gpt-4o-mini-transcribe`, `gpt-4o-transcribe-diarize` shut down Feb 26, 2027 [5]
- Live model has no timestamps or speakers: `gpt-live-transcribe` returns no word timestamps, speaker labels or confidence scores [6]

## Authentication

### Methods
| Method | Used for | How |
|---|---|---|
| API key | All API calls | Send as a bearer token in the `Authorization` header |
| Workload identity federation token | Short-lived access instead of a key | Bearer access token created by workload identity federation |
Source [9]

### Getting access
- 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 [9]
- Other requirements: Free tier requires an allowed geography [7]

## Setup

### Addresses
- REST API: `https://api.openai.com/v1/audio/transcriptions` [1]
- MCP server: None found [13]
- OpenAPI spec: Not published
- Docs for agents: `https://developers.openai.com/llms.txt`; each docs page has a `.md` copy [1]
- Status page: `https://status.openai.com` [13]

### Install
| What | Command |
|---|---|
| JavaScript SDK | `npm install openai` |
| Python SDK | `pip install openai` |
| C# SDK | `dotnet add package OpenAI` |
| Go SDK | `github.com/openai/openai-go/v3` (imported in the docs examples) |
| CLI | `openai audio:transcriptions create --model gpt-transcribe --file /path/to/file/audio.mp3` |
Source [10]

## Limits

### Rate limits
Limits shown for `gpt-transcribe`.
| Tier | Requests per minute | Tokens per minute |
|---|---|---|
| Build | 5,000 | 2,000,000 |
| Launch | 10,000 | 10,000,000 |
| Grow | 30,000 | 150,000,000 |
Source [3]

### Other limits
- File size: 25 MB per upload; split larger recordings [1]
- Diarization chunking: For audio over 30 seconds with `gpt-4o-transcribe-diarize`, set `chunking_strategy` [1]
- Known speaker references: Up to 4 reference clips of 2 to 10 seconds [1]
- Monthly usage limits: Build $500, Launch $5,000, Grow $200,000 a month [7]

## Privacy

### Your data
- Data location: Regional storage and processing in the US, Europe and other regions for `/v1/audio/transcriptions` [8]
- Retention: No abuse monitoring retention and no application state for `/v1/audio/transcriptions` [8]
- Used for training: No, unless you opt in [8]
- Zero data retention: Eligible for `/v1/audio/transcriptions` [8]

### Compliance
- Certifications: Not published
- HIPAA: Not published

## Sources

1. File transcription guide: https://developers.openai.com/api/docs/guides/speech-to-text.md
2. Transcription overview: https://developers.openai.com/api/docs/guides/transcription.md
3. GPT-Transcribe model page: https://developers.openai.com/api/docs/models/gpt-transcribe.md
4. Pricing: https://developers.openai.com/api/docs/pricing.md
5. Deprecations: https://developers.openai.com/api/docs/deprecations.md
6. Realtime transcription guide: https://developers.openai.com/api/docs/guides/realtime-transcription.md
7. Rate limits: https://developers.openai.com/api/docs/guides/rate-limits.md
8. Data controls: https://developers.openai.com/api/docs/guides/your-data.md
9. API authentication: https://developers.openai.com/api/docs/api-reference/authentication.md
10. Libraries: https://developers.openai.com/api/docs/libraries.md
11. llms.txt: https://developers.openai.com/llms.txt
12. Trust portal: https://trust.openai.com
13. Checked by supermarket.dev