Skip to main content
Audio transcription and translation use multipart uploads. Text-to-speech does not upload a file, but it does let you choose an output audio format.

​
Transcription and translation uploads

For POST /v1/audio/transcriptions and POST /v1/audio/translations, the request body is multipart/form-data.

​
Required and optional fields

FieldRequiredNotes
modelYesThe transcription or translation model
fileYesBinary audio file upload
promptNoShort hint for formatting or name preservation
languageNoOptional language hint

​
Example upload

from pathlib import Path
import requests

with Path("sample.mp3").open("rb") as audio_file:
    response = requests.post(
        "https://api.naga.ac/v1/audio/transcriptions",
        headers={"Authorization": "Bearer YOUR_API_KEY"},
        data={"model": "whisper-1", "language": "en"},
        files={"file": audio_file},
    )

response.raise_for_status()
print(response.json()["text"])

​
Validation and decoding

The gateway attempts to decode the uploaded audio before processing it. If the file cannot be decoded, the API returns an invalid_request_error. The OpenAPI contract requires a binary file field, but it does not publish a strict file-extension allowlist here. In practice, use standard decodable audio files and validate them in your own pipeline before upload.

​
Text-to-speech output formats

For POST /v1/audio/speech, the response_format field supports:
  • mp3
  • opus
  • aac
  • flac
  • wav
  • pcm

​
Practical advice

  • use clean source files whenever possible
  • prefer standard, decodable audio containers and codecs
  • keep prompts short and specific when you need name preservation or formatting hints