CONFIGURATIONS
Browse configurations by computer and inspect settings before using them. Share configuration files, not meeting data.
Configuration sharing, speaker tools and acceleration describe the current development build. See Downloads for released installer versions; older versions may not support app links.
Default 1.7B writer alongside everyday apps. Automatic backend selection checks runtime capability and available memory.
Unmeasured; a suggested combination is not a performance guarantee.
Requires a compatible MLX runtime and separate MLX weights. No published speed measurement; not the default recommendation.
Unmeasured; a suggested combination is not a performance guarantee.
A 4B model and longer context for deliberate after-meeting work. More memory does not guarantee higher speed; consider current load.
Unmeasured; a suggested combination is not a performance guarantee.
For meetings alongside other apps. Small model, short context, unloaded after each task. A suggested starting point, not a measured result.
Unmeasured; a suggested combination is not a performance guarantee.
Qwen3 1.7B generation is measured. ASR and concurrent ASR + LLM load are unmeasured. Results apply to this M5 test machine only.
Measurement included
Default 1.7B model with two threads. Memory alone is not a speed estimate; older and low-power CPUs need their own measurements.
Unmeasured; a suggested combination is not a performance guarantee.
A conservative setup for Windows, Linux and Mac. Smaller models use fewer resources but may produce weaker summaries.
Unmeasured; a suggested combination is not a performance guarantee.
No matching configurations yet. Adjust the filters or contribute your own results.
Contains settings only, no meeting content. The app will ask you to review and confirm. Missing models are not downloaded automatically.
Requires a build supporting configuration links. If the app does not open, download the file and import it in settings.