Record, transcribe and summarize meetings on your own machine. No cloud, no account, no telemetry. Built for Hebrew and mixed Hebrew/English meetings, works in 50+ languages. Speakers are told apart by voice, each with its own avatar.

   

_{Latest build from main, refreshed by CI on every push. Also available: Windows .msi · Linux .deb · versioned releases}

How language routing works

In Settings → Language map each language to its transcription and summary models. On Auto Detect, Tamlil identifies the language from the first seconds of speech and switches models for the rest of the meeting; routed languages get summaries written directly in that language. Any whisper.cpp .bin or GGUF model can be added from a URL or a file.

• Platforms: macOS on Apple Silicon (primary), Windows x64, Linux x64 (experimental). Unsigned builds: on macOS right-click → Open on first launch.
• Build from source: cd frontend && pnpm install && ./clean_run.sh. Details and architecture in docs/UPSTREAM_README.md.
• CI and releases: every push tests, bundles and refreshes the latest pre-release; signed versioned releases are manual. See docs/RELEASING.md.
• Reset or uninstall: Settings → Preferences → Storage, reset and uninstall shows everything stored on the machine and can reset it (keeping the models or not) or uninstall the app completely. If the app is already gone, run scripts/uninstall-macos.sh, scripts/uninstall-windows.ps1 or scripts/uninstall-linux.sh (--keep-models keeps the downloaded models).
• Upstream: fork of Meetily (MIT). Fork code lives in model_registry/ and language_routing/; sync with git fetch upstream main && git rebase upstream/main.

Credits

Tamlil is built on Meetily by Zackriya Solutions and its contributors, released under the MIT License (see LICENSE.md). Meetily in turn borrows from whisper.cpp, Screenpipe and transcribe-rs, and uses NVIDIA's Parakeet model in the ONNX conversion by istupakov. Hebrew models come from ivrit.ai; speaker embeddings from WeSpeaker. The original README is kept at docs/UPSTREAM_README.md.