Speech-to-Text API

Transcribe audio files into text with state-of-the-art accuracy. Supports 99+ languages, speaker diarization, word-level timestamps, and translation to English. Transcribes 30 minutes of audio in under one minute.

Endpoint

POST https://api.tryspeakeasy.io/v1/audio/transcriptions

Authentication

All requests require a Bearer token in the Authorization header. You can generate API keys from your dashboard.

Authorization: Bearer YOUR_API_KEY

Request Parameters

The request body must be sent as multipart/form-data.

Code Examples

cURL

curl -X POST https://api.tryspeakeasy.io/v1/audio/transcriptions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@recording.mp3" \
  -F "model=whisper-large-v3" \
  -F "response_format=verbose_json" \
  -F "speaker_labels=true" \
  -F "word_timestamps=true" \
  -F "language=english"

Python (OpenAI SDK)

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.tryspeakeasy.io/v1"
)

# Basic transcription
transcript = client.audio.transcriptions.create(
    model="whisper-large-v3",
    file=open("recording.mp3", "rb"),
    response_format="verbose_json"
)

print(transcript.text)
print(f"Duration: {transcript.duration}s")
print(f"Language: {transcript.language}")

JavaScript

const formData = new FormData();
formData.append('file', audioFile);
formData.append('model', 'whisper-large-v3');
formData.append('response_format', 'verbose_json');
formData.append('speaker_labels', 'true');
formData.append('word_timestamps', 'true');

const response = await fetch('https://api.tryspeakeasy.io/v1/audio/transcriptions', {
  method: 'POST',
  headers: {
    Authorization: `Bearer ${apiKey}`,
  },
  body: formData,
});

const result = await response.json();
console.log(result.text);
console.log(result.segments);
console.log(result.words);

Transcription via URL

Instead of uploading a file directly, you can pass a publicly accessible URL pointing to the audio. This is especially useful for large files — URL-based transcription supports files up to 1 GB (compared to 25 MB for direct uploads).

curl -X POST https://api.tryspeakeasy.io/v1/audio/transcriptions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "url=https://example.com/files/meeting-recording.mp3" \
  -F "model=whisper-large-v3" \
  -F "response_format=verbose_json" \
  -F "speaker_labels=true"

Async Transcription (Callback URL)

For long audio files that take a while to transcribe, you can provide a callback_url instead of waiting for the API to finish processing. The API returns 202 Accepted immediately, then sends a POST request to your callback URL with the transcription result when it is ready.

This frees your client from holding a connection open and avoids timeouts on very long recordings.

Example Request

curl -X POST https://api.tryspeakeasy.io/v1/audio/transcriptions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@long-recording.mp3" \
  -F "model=whisper-large-v3" \
  -F "response_format=verbose_json" \
  -F "callback_url=https://your-server.com/webhooks/transcription"

Callback Payload

When processing is complete, the API sends a POST request to your callback URL. The request body contains the transcription result in the same format as a synchronous response:

{
  "text": "Hello, welcome to the meeting...",
  "language": "en",
  "duration": 1823.45,
  "segments": [
    {
      "id": 0,
      "start": 0.0,
      "end": 2.16,
      "text": "Hello, welcome to the meeting."
    }
  ]
}

Using the Prompt Parameter

The prompt parameter lets you guide the transcription model. It is particularly useful when your audio contains domain-specific terms or when you want to control the output style. The prompt should be in the same language as the audio.

Fix domain-specific terms and acronyms

If the model misrecognizes specialized terms, provide them in the prompt:

-F "prompt=NFT, DeFi, DAO, HODL, Ethereum, Solana"

Preserve filler words

By default the model may omit filler words. Include them in the prompt to keep them in the output:

-F "prompt=Um, uh, like, you know"

Continue a previous transcript

When transcribing audio in chunks, pass the end of the previous transcript as the prompt. This helps the model maintain context and consistent punctuation:

-F "prompt=...and that concludes our discussion on the quarterly results."

Response Schema

When response_format is set to verbose_json, the response includes the full transcription result with segments, word-level timestamps, and speaker labels (if enabled).

Example Response

A verbose_json response with word_timestamps and speaker_labels enabled:

{
  "text": "Hello, welcome to the meeting. Thank you for joining us today.",
  "language": "en",
  "duration": 5.42,
  "segments": [
    {
      "id": 0,
      "start": 0.0,
      "end": 2.16,
      "text": "Hello, welcome to the meeting."
    },
    {
      "id": 1,
      "start": 2.48,
      "end": 5.42,
      "text": "Thank you for joining us today."
    }
  ],
  "words": [
    { "word": "Hello,", "start": 0.0, "end": 0.42 },
    { "word": "welcome", "start": 0.44, "end": 0.82 },
    { "word": "to", "start": 0.84, "end": 0.96 },
    { "word": "the", "start": 0.98, "end": 1.08 },
    { "word": "meeting.", "start": 1.10, "end": 1.62 },
    { "word": "Thank", "start": 2.48, "end": 2.78 },
    { "word": "you", "start": 2.80, "end": 2.96 },
    { "word": "for", "start": 2.98, "end": 3.14 },
    { "word": "joining", "start": 3.16, "end": 3.58 },
    { "word": "us", "start": 3.60, "end": 3.78 },
    { "word": "today.", "start": 3.80, "end": 4.32 }
  ],
  "speakers": [
    {
      "speaker": "SPEAKER_00",
      "start": 0.0,
      "end": 2.16,
      "text": "Hello, welcome to the meeting."
    },
    {
      "speaker": "SPEAKER_01",
      "start": 2.48,
      "end": 5.42,
      "text": "Thank you for joining us today."
    }
  ]
}

Supported Audio Formats

The following audio and video formats are accepted:

Supported Languages

Pass one of the following values as the language parameter. 99 languages are supported:

Error Responses

The API returns standard HTTP error codes with a JSON body describing the error. Common errors for this endpoint include:

For a full list of error codes and troubleshooting guidance, see the Error Codes reference.