Audio Transcriptions API

Reference for POST /v1/audio/transcriptions, including file uploads, file URLs, language hints, response formats, and billing.

This endpoint is implemented by the UnderSky Rust gateway. Availability depends on the configured upstream provider and media storage.

Transcribe audio into text. Supports file upload or URL.

Endpoint

POST https://api.undersky.ai/v1/audio/transcriptions

Models

Model	Pricing	Description
`rapid`	0.006 credits/min	Fast transcription, supports file_uri
`whisper-1`	0.006 credits/min	High accuracy with prompt support

Parameters

Parameter	Type	Required	Description
file	file	Either `file` or `file_uri`	Audio file to transcribe (multipart upload)
file_uri	string	Either `file` or `file_uri`	URL of the audio file to transcribe
model	string	No	Model to use: `"rapid"` or `"whisper-1"`. Default: `"rapid"`
response_format	string	No	Output format: `"text"`, `"json"`, `"verbose_json"`, `"srt"`, or `"vtt"`. Default: `"json"`
language	string	No	ISO 639-1 language code (e.g. `"en"`, `"zh"`). Improves accuracy if specified
prompt	string	No	Hint text to improve transcription quality
temperature	number	No	Sampling temperature between 0 and 1 (`whisper-1` only)

Request

Upload a file

curl -X POST https://api.undersky.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer sk-your-api-key" \
  -F [email protected] \
  -F model=rapid \
  -F response_format=json

Pass a URL (recommended for large files)

curl -X POST https://api.undersky.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer sk-your-api-key" \
  -F file_uri=https://example.com/audio.mp3 \
  -F model=rapid \
  -F language=zh \
  -F response_format=verbose_json

Using file_uri avoids uploading large files through your network. The audio is fetched server-side. Supported with both rapid and whisper-1 models.

Using whisper-1 with prompt

curl -X POST https://api.undersky.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer sk-your-api-key" \
  -F [email protected] \
  -F model=whisper-1 \
  -F response_format=json \
  -F prompt="UnderSky, GTC, NVIDIA"

Requesting subtitle output

curl -X POST https://api.undersky.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer sk-your-api-key" \
  -F [email protected] \
  -F model=whisper-1 \
  -F response_format=srt

Response

{
  "text": "Hello, this is a sample transcription of the audio file."
}

When using verbose_json, the response includes timestamps and segments:

{
  "task": "transcribe",
  "language": "en",
  "duration": 42.5,
  "text": "Hello, this is a sample transcription of the audio file.",
  "segments": [
    { "start": 0.0, "end": 2.4, "text": "Hello, this is a sample transcription." }
  ]
}

When using srt, the response is plain text subtitle output:

1
00:00:00,000 --> 00:00:02,400
Hello, this is a sample transcription.

When using vtt, the response is returned as WebVTT:

WEBVTT

00:00:00.000 --> 00:00:02.400
Hello, this is a sample transcription.

Supported Audio Formats

mp3, mp4, mpeg, mpga, m4a, wav, webm, opus

File uploads are limited to 25 MB. When using file_uri, there is no size limit.

Notes

Billing is based on audio duration, charged at the per-model rate listed above.
If no model is specified, rapid is used by default.
Streaming (stream=true) is not supported at this time.

Audio Transcriptions API

On this page