Direct streaming. More reliable dictation.
AirNote 2.4.5 is the current stable macOS release. Polished dictation now reaches your focused app as it streams, while the desktop app and server safely recover from interrupted output. The current Windows release is 2.4.4.
See your dictation arrive as it is polished
AirNote now sends polished dictation into the focused app as the provider streams it, so the result no longer waits for a separate final insertion step.
If a stream is interrupted or its final result differs, AirNote reconciles the recording safely instead of duplicating the text.
You dictated 39 words today.
That's roughly 1 minute saved vs. typing, at an average of 155 WPM.
- Direct streamed insertion
- No duplicate final paste
- Safe interrupted-stream recovery
A steadier speech-to-text path
The desktop speech path uses the current DeepInfra Whisper integration, with resilient handling for slow responses, retries, and recovery after an app restart.
AirNote keeps your normal dictation path focused on the model’s output while preserving the meeting and background-processing workflows separately.
Today · Jun 17
Ready when you are.
Hold your hotkey and speak — AirNote types polished text into the focused app.
At a glance
Activity — last 14 days
Today — 0 recordings
Your first dictation will land here.
Today — 1 recording
- Improved speech-service resilience
- Clearer failure recovery
- Separate meeting processing
Clearer operational insight
Meeting telemetry now maintains durable, ordered delivery so activity remains correctly attributed across workspaces and recovers cleanly after interruptions.
The admin experience makes dictation and meeting usage easier to inspect, including explicit no-AI-usage and retryable-error states.
Today · Jun 17
Ready when you are.
Hold your hotkey and speak — AirNote types polished text into the focused app.
At a glance
Activity — last 14 days
Today — 0 recordings
Your first dictation will land here.
Today — 1 recording
- Durable meeting telemetry
- Cross-workspace attribution protection
- Clear admin reporting states
Detailed changes
Direct streaming3
- Polished text now arrives in the focused app as it streams.
- Removed the delayed post-model insertion path for normal dictation.
- Added safe reconciliation for interrupted or divergent streams.
Reliability3
- Improved recovery from slow or interrupted speech-service responses.
- Kept normal dictation model output direct and predictable.
- Preserved separate processing for meetings and background services.
Meetings & visibility3
- Protected cross-workspace meeting telemetry attribution.
- Kept blocked events from starving the outbox.
- Added clearer admin usage and retryable-error states.
Current stable downloads
Latest stable for macOS
Current stable for Windows