Transcription

Casola transcribes spoken audio into text with timestamps and segment data. Navigate to /transcribe in Studio to get started.

Providing audio

Three input methods are available:

Upload a file — Drag and drop an audio file onto the upload area, or click to browse. Supports all standard audio formats (MP3, WAV, M4A, OGG, etc.) up to 25 MB.

Record audio — Click the record button to capture audio directly from your microphone. A playback preview appears when you stop recording so you can verify before submitting.

Audio URL — Paste a direct HTTPS link to an audio file hosted online.

Settings

Model — Casola currently supports Whisper Large v3 for speech-to-text. The model selector shows availability status. See the Models reference for the latest options.

Language — Choose the spoken language or leave on Auto-detect (the default). Specifying the language can improve accuracy for non-English audio. Supported languages include English, Spanish, French, German, Italian, Portuguese, Russian, Japanese, Korean, and Chinese.

Response format — Select how the transcription is structured:

Format	Description	Best for
JSON	Full response with segments, timestamps, and metadata	Programmatic use
Text	Plain text transcription	Quick reading, copy-paste
SRT	SubRip subtitle format with timestamps	Video subtitles
VTT	WebVTT subtitle format with timestamps	Web video players

Reading results

Completed transcriptions display the full text along with metadata: word count, character count, detected language, and audio duration.

When the model returns segments (available with JSON format), each segment is shown with its start and end timestamps — useful for aligning text to specific moments in the audio.

You can toggle between a formatted text view and a raw JSON view when using the JSON format.

Exporting results

• Copy — Click the copy button to place the full transcription on your clipboard
• Download — Save the transcription as a file matching your chosen format (.json, .txt, .srt, or .vtt)

All transcription results are automatically saved to your Library.

API usage

Transcribe from a URL

curl https://api.casola.ai/openai/v1/audio/transcriptions \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "whisper-large-v3",
    "audio_url": "https://example.com/meeting-recording.mp3",
    "language": "en",
    "response_format": "verbose_json"
  }'

Response:

{
  "task": "transcribe",
  "language": "en",
  "duration": 125.4,
  "text": "Welcome to today's meeting. Let's start with the agenda...",
  "segments": [
    {
      "start": 0.0,
      "end": 3.2,
      "text": "Welcome to today's meeting."
    },
    {
      "start": 3.5,
      "end": 6.1,
      "text": "Let's start with the agenda..."
    }
  ]
}

Upload a file

Use multipart/form-data to upload an audio file directly:

curl https://api.casola.ai/openai/v1/audio/transcriptions \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -F model="whisper-large-v3" \
  -F file=@recording.mp3 \
  -F response_format="json"

Response:

{
  "text": "Welcome to today's meeting. Let's start with the agenda..."
}

Get SRT subtitles

curl https://api.casola.ai/openai/v1/audio/transcriptions \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "whisper-large-v3",
    "audio_url": "https://example.com/video-audio.mp3",
    "response_format": "srt"
  }'

Response (plain text):

1
00:00:00,000 --> 00:00:03,200
Welcome to today's meeting.

2
00:00:03,500 --> 00:00:06,100
Let's start with the agenda...

Python (OpenAI SDK)

from openai import OpenAI

client = OpenAI(
    base_url="https://api.casola.ai/openai/v1",
    api_key="YOUR_API_TOKEN",
)

# From file
with open("recording.mp3", "rb") as f:
    transcript = client.audio.transcriptions.create(
        model="whisper-large-v3",
        file=f,
        response_format="verbose_json",
    )
print(transcript.text)
for segment in transcript.segments:
    print(f"[{segment.start:.1f}s] {segment.text}")

Processing time

Transcription speed depends on the length of the audio. Most files under 5 minutes complete within 10–30 seconds. Longer recordings may take proportionally more time.

Tips

• For best accuracy, use audio with clear speech and minimal background noise.
• Specify the language explicitly when transcribing non-English audio — auto-detection works well but a language hint improves results.
• Use SRT or VTT format if you need subtitles for a video project.
• The JSON format includes the richest data (segments with timestamps) and is the best choice when you need to process the output programmatically.