User context
Hi, and thank you for making OpenLess open source.
I mainly use the Toggle recording mode: press once to start speaking, then press again to stop and submit. It is much more comfortable than holding a key, but having to press the hotkey a second time still interrupts my flow when I am repeatedly talking to AI, drafting prompts, or organizing longer thoughts.
I would really appreciate an optional “stop when I finish speaking” feature. After recording starts and actual speech has been detected, a sustained period of silence would automatically end the recording and continue through the existing transcription, polishing, and insertion flow. The interaction would become: press once, speak naturally, and wait for the result.
Suggested behavior
- Apply only to Toggle mode and keep it disabled by default, so existing behavior does not change.
- Start the silence timer only after real speech has been detected.
- Automatically stop and submit after continuous silence reaches a configurable threshold.
- Offer practical choices such as 1, 1.5, 2, 3, 4, and 5 seconds; 3 seconds seems like a safe default.
- If no speech is detected at all (for example, within 10 seconds), cancel instead of submitting an empty recording.
- Keep the second hotkey press as a manual early-stop action, and keep Esc as cancel.
- Leave Push-to-talk behavior unchanged.
- Adapt to microphone noise levels where possible, so a start sound, keyboard tap, or short background noise does not count as speech.
Why this would be useful
OpenLess is especially useful for writing AI prompts, asking questions, and dictating into arbitrary text fields. Silence-based auto-stop would make it feel much closer to natural hands-free voice input instead of a recorder that always needs two key presses. It would also help people who dictate frequently or find repeated key presses inconvenient.
Backward compatibility and maintenance
To make future upstream updates easier, the silence detector could remain a small independent component that produces a one-shot stop/cancel decision from the existing audio level stream or a VAD. The recording state machine would only consume that decision. New preference fields could have backward-compatible defaults, with the feature disabled when older configuration files do not contain them.
Thank you for considering this. If the direction fits the project, I would be happy to provide real-world feedback on silence timings and different background-noise conditions.
User context
Hi, and thank you for making OpenLess open source.
I mainly use the Toggle recording mode: press once to start speaking, then press again to stop and submit. It is much more comfortable than holding a key, but having to press the hotkey a second time still interrupts my flow when I am repeatedly talking to AI, drafting prompts, or organizing longer thoughts.
I would really appreciate an optional “stop when I finish speaking” feature. After recording starts and actual speech has been detected, a sustained period of silence would automatically end the recording and continue through the existing transcription, polishing, and insertion flow. The interaction would become: press once, speak naturally, and wait for the result.
Suggested behavior
Why this would be useful
OpenLess is especially useful for writing AI prompts, asking questions, and dictating into arbitrary text fields. Silence-based auto-stop would make it feel much closer to natural hands-free voice input instead of a recorder that always needs two key presses. It would also help people who dictate frequently or find repeated key presses inconvenient.
Backward compatibility and maintenance
To make future upstream updates easier, the silence detector could remain a small independent component that produces a one-shot stop/cancel decision from the existing audio level stream or a VAD. The recording state machine would only consume that decision. New preference fields could have backward-compatible defaults, with the feature disabled when older configuration files do not contain them.
Thank you for considering this. If the direction fits the project, I would be happy to provide real-world feedback on silence timings and different background-noise conditions.