Quelque is an open-source tool designed to transform long-form educational audio (lectures, class recordings, study sessions) into structured, highly readable study materials.
Rather than just providing a raw transcript, Quelque uses a two-step pipeline:
- Transcription: Converts the audio to a highly accurate, word-for-word transcript.
- LLM Sanitization & Summarization: Analyzes the raw transcript to clean speech-to-text errors and punctuation, and synthesizes it into a concise summary, key takeaways, detailed study notes, a glossary of terms, and actionable items.
- Robust Processing: Upload MP3, WAV, or M4A files. The audio is automatically normalized, chunked, and transcribed.
- Multiple AI Providers: Choose your preferred engine for both steps:
- OpenAI: High-quality transcription (Whisper) and summarization (GPT-4o) using modern cloud models.
- Google Gemini: Lightning-fast, high-quality notes generation using Google AI Studio.
- No Host (Free): 100% zero-cost processing using a local
faster-whispermodel for offline transcription and the Hugging Face Free Serverless API (Qwen-2.5-7B) for summarization.
- Language Hints: Force the transcriber to listen in a specific language and instruct the AI to write the final notes in that same language.
- Rich Exports: Download your generated notes in Markdown, DOCX (Microsoft Word), or PDF formats.
- Smart Caching: Audio processing is cached using SHA-256 hashing to prevent redundant API calls and save time.
Ensure you have Python 3.11+ and ffmpeg installed on your system.
# Clone the repository
git clone https://github.com/your-username/quelque.git
cd quelque
# Create and activate a virtual environment
python -m venv .venv
# Windows: .venv\Scripts\activate
# Unix/MacOS: source .venv/bin/activate
# Install the application and dependencies
pip install -e .
# Set up environment variables
copy .env.example .envIf you plan to use the hosted cloud models, edit your .env file to include your OPENAI_API_KEY, GOOGLE_API_KEY, or HF_TOKEN.
Start the web interface:
streamlit run streamlit_app.pyQuelque also includes a powerful CLI for batch processing:
# Generate full notes using the default local transcription and Hugging Face API
quelque notes samples/tiny_sanitized_sample.wav --language en
# Transcribe audio using OpenAI instead of the local provider
quelque transcribe samples/tiny_sanitized_sample.wav --provider openai --language enQuelque is built with a modular, API-first design that separates audio processing, transcription, and text analysis.
flowchart LR
A[Audio Upload] --> B[Normalize + Chunk]
B --> C[Transcription Provider]
C --> D[Transcript Merge]
D --> E[Notes Generation Provider]
E --> F[Cache + Exports]
F --> G[Streamlit UI / CLI]
By default, Quelque is configured to use:
- OpenAI Transcription:
gpt-4o-mini-transcribe - OpenAI Notes:
gpt-4o-mini - Google Notes:
gemini-2.5-flash - No Host Notes:
Qwen/Qwen2.5-7B-Instruct(via Hugging Face API) - Local Transcription:
faster-whisper(small model)
You can override these defaults by setting the corresponding environment variables (e.g., QUELQUE_LOCAL_TRANSCRIPTION_MODEL=base).
- Ephemeral Keys: When running the UI, user-provided API keys are kept strictly in memory for that specific session and are explicitly masked in internal configurations to prevent leakage in logs or tracebacks.
- Hugging Face Spaces: Quelque is fully Dockerized and ready to be deployed to a Hugging Face Docker Space. Simply create a new Space, push the repository, and it will work out of the box.
MIT License. See LICENSE for details.