# whisper-1

> OpenAI whisper-1 speech-to-text through the LLM gateway with the OpenAI SDK (POST /llm/openai/v1/audio/transcriptions, multipart) - json or verbose_json with segment/word timestamps. Billed per second of audio (0.1858 credits/s).

> **Retirement notice**: the upstream provider retires this model on 2027-02-26. Requests keep working until then — migrate before that date.

- **Provider**: openai
- **Model ID**: whisper-1
- **Category**: speech_to_text
- **Pricing Type**: token_based
- **Billing**: per second of input audio — see Pricing

## API Endpoint

Base URL: https://api.core.today

### Audio Transcription
POST /llm/openai/v1/audio/transcriptions

## Authentication

Header: `Authorization: Bearer YOUR_API_KEY`

Note: Your Core.Today API key (`cdt_xxx`) works as Bearer token.

## Input Parameters

- Send `multipart/form-data` (the OpenAI SDK's `client.audio.transcriptions.create`), not JSON
- `file` (file, **required**): The audio file — flac, m4a, mp3, mp4, mpeg, mpga, ogg, wav or webm, max 25MB
- `model` (string, **required**): Model identifier. Use `whisper-1`
- `response_format` (string, optional): `json` or `verbose_json` (Default: `json`). `text`, `srt` and `vtt` are rejected with 400 — build subtitles from verbose_json segments
- `timestamp_granularities[]` (array, optional): `segment` and/or `word`; requires `response_format=verbose_json`
- `language` (string, optional): ISO-639-1 code of the audio (e.g. `ko`, `en`)
- `prompt` (string, optional): Text to guide style or continue a previous segment
- `temperature` (number, optional): Sampling temperature (Default: `0`; Range: min: 0, max: 1)

## Example Request

Multipart form fields:

```json
{
  "file": "@meeting.mp3",
  "model": "whisper-1",
  "response_format": "verbose_json",
  "timestamp_granularities[]": "segment"
}
```

```bash
curl https://api.core.today/llm/openai/v1/audio/transcriptions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F file=@meeting.mp3 \
  -F model=whisper-1 \
  -F response_format=verbose_json \
  -F "timestamp_granularities[]=segment"
```

OpenAI SDK: `OpenAI(base_url="https://api.core.today/llm/openai/v1", api_key="cdt_...")` then `client.audio.transcriptions.create(...)` with the same fields.

## Response Format

```json
{
  "task": "transcribe",
  "language": "english",
  "duration": 52.41,
  "text": "The little tales they tell are false. ...",
  "segments": [
    {
      "id": 0,
      "start": 0.0,
      "end": 5.04,
      "text": " The little tales they tell are false."
    }
  ],
  "usage": {
    "type": "duration",
    "seconds": 53
  }
}
```

## Usage Flow

1. POST /llm/openai/v1/audio/transcriptions with a multipart/form-data upload of file and model
2. The transcript is returned synchronously; verbose_json adds language, duration and timestamped segments/words
3. Credits are deducted per second of audio, on the length OpenAI reports (`usage.seconds` / `duration`)

Credits in the response (opt-in): send the header `X-Coredot-Include-Cost: true` and the response reports the credits that call is billed — `usage.coredot.credits` in the body (Gemini: `usageMetadata.coredot`; streaming: in the final usage event) and, for non-streaming responses, the `X-Coredot-Credits` header. Guide: https://console.core.today/llms/api.txt (section "Credits in the response").

## Pricing

- Audio: 0.1858 credits per second (≈ 11.148 credits per minute)

## Tags

openai, speech_to_text