AuraFlow
Keep the words from the conversation.
AuraFlow turns speech into text and helps prepare meeting notes. Speech recognition and language models run locally.
Let us send the draft on Friday. Check the scope first.
Start with what was said.
The recording preserves context. The transcript lets you return to a specific part of the conversation. A model helps prepare a summary, which you then check.
One conversation. A readable record.
This example follows a spoken sentence into draft notes. Compare the note with what was said.
Spoken words
Let us send the draft on Friday. Check the scope first.
Transcript
Let us send the draft on Friday. Check the scope first.
Draft notes
Plan: send the draft on Friday. Check the scope before sending it.
Check decisions and commitments against the transcript before you share the notes.
The models run on your device.
Speech recognition and summaries do not require a cloud model. The initial model download needs a network connection. If you enable export to an external system, data goes to the destination you choose.
Choose a device. Check the scope.
Features vary by device. The macOS and Android releases are available. The Windows version is under device testing.
- macOS
- Apple Silicon release
Local transcription, summaries and note export.
- Android
- Native release
Speech recognition and a language model on the phone.
- Windows
- On-device acceptance
The core process passed tests. Full interface and microphone acceptance remains open.
Notes ready for your review.
Describe how you work with conversations and which device you use. We can identify the AuraFlow variant that fits your process.
An Aurora AI project. The streaming engine is adapted from whisper-flow, and desktop application patterns from OpenWhispr. The AuraFlow family includes a meeting variant.