Speech to Text

CompactifAI’s speech-to-text capability delivers fast, reliable transcripts from common audio formats while staying fully compatible with the OpenAI Whisper API surface. The endpoint accepts multipart form uploads and returns JSON, plain text, or verbose JSON depending on response_format. It is ideal for meeting notes, support calls, or media captioning.

Basic Usage (Python)

import requests

API_URL = "https://api.compactif.ai/v1/audio/transcriptions"
API_KEY = "your_api_key_here"

headers = {
    "Authorization": f"Bearer {API_KEY}"
}

payload = {
    "model": "whisper-large-v3",
    "language": "en",
    "temperature": 0,
    "response_format": "json"
}
file_name = "meeting_minutes.mp3"
file_content_type = "audio/mpeg"
with open(file_name, "rb") as audio_file:
    response = requests.post(API_URL, headers=headers, data=payload, files={"file": (file_name, audio_file, file_content_type)})

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

Accepted Parameters

Field	Type	Description
file	file upload	Supported files types - .flac, .mp3, .mp4, .mpga, .m4a, .wav, .webm, .mpeg, and .ogg (Note - For .ogg and .mpeg, the system only supports audio files)
model	string	Required model alias such as whisper-large-v3
prompt	string	Optional text primer to bias the transcription
temperature	number	Optional float between 0 and 1 (defaults to provider setting)
language	string	Optional ISO language code hint (en by default)
response_format	string	Output format: json (default), text, or verbose_json. json and verbose_json return JSON; text returns plain text (Content-Type: text/plain).
stream	boolean	Whether to stream back partial progress

Example Response (response_format=json)

{
  "text": "Welcome to the quarterly planning meeting. Let's review the agenda.",
  "logprobs": null,
  "usage": {
    "type": "duration",
    "seconds": 12.6
  }
}

With response_format=text, the response body is the transcript string only (Content-Type: text/plain). Use response_format=verbose_json when you need additional metadata such as duration, language, segments, and words.

Tips

• Keep uploads under 25 MB for best performance; large files benefit from client-side compression.
• Provide the language hint when you know the spoken language to reduce warm-up time.
• Use the same headers and authentication flow as other CompactifAI endpoints; only the form payload differs.