The 30-second recording cap is an inherent Whisper constraint. For long-form dictation use cases, I use voice-to-text for clinical documentation, manually retriggering the record button every cycle is a significant workflow disruption.
Proposal: add a toggle mode that auto-restarts recording after each chunk is transcribed. Record for ~28 seconds, stop, transcribe and insert text, immediately begin the next recording, looping until the user disables the toggle. This mirrors how desktop Whisper handles long-form audio via sliding windows, just adapted to the IME interaction model.
I've hacked together a workaround using Tasker + AutoInput to simulate the button press on a timer, which tells me the workflow is viable, it just needs a native implementation. This would make the app substantially more useful for professional dictation without requiring any changes to the underlying model. Thanks for the excellent work on this project. Warm regards.
The 30-second recording cap is an inherent Whisper constraint. For long-form dictation use cases, I use voice-to-text for clinical documentation, manually retriggering the record button every cycle is a significant workflow disruption.
Proposal: add a toggle mode that auto-restarts recording after each chunk is transcribed. Record for ~28 seconds, stop, transcribe and insert text, immediately begin the next recording, looping until the user disables the toggle. This mirrors how desktop Whisper handles long-form audio via sliding windows, just adapted to the IME interaction model.
I've hacked together a workaround using Tasker + AutoInput to simulate the button press on a timer, which tells me the workflow is viable, it just needs a native implementation. This would make the app substantially more useful for professional dictation without requiring any changes to the underlying model. Thanks for the excellent work on this project. Warm regards.