Groq Transcriber

A lightweight web app for transcribing audio with Groq Whisper models. Upload a file, pick a model and timestamp detail level, and download the result as plain text, JSON, or Markdown.

Features

• Whisper models — whisper-large-v3-turbo (fast, $0.04/hr) and whisper-large-v3 (highest accuracy, $0.111/hr)
• Timestamps detail — sentences (default), paragraphs, plain text, or raw Whisper segments
• Export formats — TXT, JSON (includes raw segments + formatted blocks), Markdown
• Cost estimate — per-run billing estimate based on audio duration
• Progress UI — step timeline, upload progress, elapsed time
• Docker ready — single-container deploy with persisted API key storage

Quick start

Local

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

CONFIG_DIR=./data uvicorn app.main:app --host 127.0.0.1 --port 8080 --reload

Open http://127.0.0.1:8080, add your Groq API key in Settings, then transcribe.

Docker

docker compose up --build

The app listens on port 8080. Your API key is stored in the transcriber-data volume.

Configuration

Supported audio

MP3, MP4, MPEG, MPGA, M4A, WAV, WebM, OGG, FLAC — up to 100 MB per file.

License

MIT — see LICENSE.