Transform lecture audio into structured learning materials — completely private, fully offline.
Knowte is a desktop application that turns lecture recordings into structured notes, interactive quizzes, flashcards, mind maps, and related research papers. All processing happens locally on your machine using Ollama for LLM inference and Whisper for transcription. No data ever leaves your device.
- 🎙️ Audio & Video Input — Upload audio/video files or record directly from your microphone. Supports MP3, WAV, M4A, OGG, WebM, MP4, MKV, and more.
- 📺 YouTube Import — Paste a YouTube URL and Knowte downloads and transcribes the audio using
yt-dlp. - 📝 Transcription — Local speech-to-text using Whisper.cpp with model sizes from tiny (~75 MB) to large (~3 GB).
- 📖 Structured Notes — LLM-generated notes organised into topics, key points, examples, key terms, and takeaways.
- ✅ Interactive Quiz — Auto-generated multiple choice, true/false, and short-answer questions with explanations and score tracking.
- 🃏 Flashcards — Anki-style flashcards with card-flip animations, three-pile study mode, and Anki
.apkg/.txtexport. - 🧠 Mind Map — Visual tree of lecture concepts using React Flow, with PNG and SVG export.
- 🔬 Research Papers — Related academic papers via the Semantic Scholar API (optional, requires internet).
- 💡 Explain This — Select any text and get a contextual AI explanation, with "simpler" / "deeper" controls.
- 📚 Lecture Library — Persistent SQLite-backed history of all lectures with full-text search.
- 🔒 100% Local & Private — No cloud API calls (except the optional Semantic Scholar paper search). Audio and transcripts never leave your machine.
Ollama runs language models locally. Download and install it from ollama.ai, then pull a model:
# Recommended — good balance of quality and speed (~4.7 GB)
ollama pull llama3.1:8b
# Lightweight option (~2.3 GB)
ollama pull phi3:miniVerify it's running:
curl http://localhost:11434/api/tagsKnowte bundles platform-specific ffmpeg and yt-dlp binaries. For development see CONTRIBUTING.md.
- Install prerequisites (Ollama + a model — see above)
- Download Knowte from the Releases page and install it
- On first launch, the setup wizard guides you through:
- Verifying Ollama is running
- Choosing a language model
- Downloading a Whisper transcription model
- Setting your academic level for personalised outputs
- Upload a lecture — drag and drop an audio/video file, or paste a YouTube URL
- Click "Process Knowte" — Knowte transcribes and generates all materials automatically
| Tool | Version | Install |
|---|---|---|
| Rust | stable | curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh |
| Bun | ≥ 1.0 | curl -fsSL https://bun.sh/install | bash |
| Tauri CLI | v2 | included in devDependencies |
| Ollama | any | ollama.ai |
git clone https://github.com/GxAditya/Knowte.git
cd knowte
bun install
bun run tauri devnpx tsc --noEmit # Type-check
bun run tauri build # Build installerBuild artefacts are in src-tauri/target/release/bundle/:
| Platform | Formats |
|---|---|
| Linux | .deb, .AppImage, .rpm |
| macOS | .dmg |
| Windows | .msi, .exe (NSIS) |
All user data — audio files, transcripts, notes, quizzes, flashcards — lives in SQLite in the platform app data directory:
| Platform | Path |
|---|---|
| Linux | ~/.local/share/com.knowte.app/ |
| macOS | ~/Library/Application Support/com.knowte.app/ |
| Windows | %APPDATA%\com.knowte.app\ |
The only optional external network call is to Semantic Scholar for related paper search (disable in Settings → Research).
GPL 3.0 License — see LICENSE for details.
See CONTRIBUTING.md for details.