feat: implement voice mode improvements with continuous loop, audio error recovery, and TTS discovery#459

Open

devin-ai-integration[bot] wants to merge 31 commits intomasterfrom

devin/1772755465-voice-mode-improvements

Commits on Mar 6, 2026

feat: implement voice mode improvements with continuous loop, audio error recovery, and TTS discovery

devin-ai-integration[bot]
and
marksftw
committed
style: fix Prettier formatting in RecordingOverlay and UnifiedChat

devin-ai-integration[bot]
and
marksftw
committed
fix: track recording start time with ref, fix stale closure in handleVoiceDiscard

devin-ai-integration[bot]
and
marksftw
committed
fix: use startRecordingRef in voice continuation effect, fix || vs ?? for savedDuration

devin-ai-integration[bot]
and
marksftw
committed
fix: add isTauri() guard for iOS TTS platform check, restart recording on empty blob

devin-ai-integration[bot]
and
marksftw
committed
fix: use refs for exitVoiceMode recorder check and stale handleSendMessage/handleTTSDiscovery

devin-ai-integration[bot]
and
marksftw
committed
style: fix Prettier formatting for ref type declarations

devin-ai-integration[bot]
and
marksftw
committed
fix: call cancelTTSGeneration/stopTTS unconditionally in exitVoiceMode to avoid stale closure

devin-ai-integration[bot]
and
marksftw
committed
fix: clear voiceState after non-voice retry success, exit voice mode on generation error

devin-ai-integration[bot]
and
marksftw
committed
fix: re-throw TTS playback error so speakAndWait rejects and voice mode exits

devin-ai-integration[bot]
and
marksftw
committed
fix: catch errors in speak() wrapper to prevent unhandled rejection in TTSButton

devin-ai-integration[bot]
and
marksftw
committed
fix: reset voiceRetryCount and recordingDuration when dismissing error overlay via X button

devin-ai-integration[bot]
and
marksftw
committed
fix: TestFlight bugs - show playing state in compact overlay, iOS audio cues, TTS auto-play after model download

devin-ai-integration[bot]
and
marksftw
committed
style: run Prettier on RecordingOverlay and UnifiedChat

devin-ai-integration[bot]
and
marksftw
committed
fix: reset timer on re-entering recording state, remove duplicate mic-on audio cue

devin-ai-integration[bot]
and
marksftw
committed
fix: prevent AudioContext leak and handle empty preprocessed TTS text

devin-ai-integration[bot]
and
marksftw
committed
fix: exit voice mode on startRecording failure, guard stale audioContextRef cleanup

devin-ai-integration[bot]
and
marksftw
committed
fix: TestFlight round 2 - iOS audio cues resume(), duration flash fix, compact layout gap, TTS auto-play after download

devin-ai-integration[bot]
and
marksftw
committed
debug: add test buttons for audio cue debugging on mobile Safari

devin-ai-integration[bot]
and
marksftw
committed
fix: cancelGeneration now stops playback to prevent audio during async decoding window

devin-ai-integration[bot]
and
marksftw
committed
fix: play audio cues before mic activation and after mic deactivation on iOS

devin-ai-integration[bot]
and
marksftw
committed
fix: restore audioSession type after cue playback to allow mic capture

devin-ai-integration[bot]
and
marksftw
committed
fix: use play-and-record audioSession type instead of playback

devin-ai-integration[bot]
and
marksftw
committed
chore: replace audio cue WAV files with louder, trimmed versions

devin-ai-integration[bot]
and
marksftw
committed
fix: remove 0.5 gain reduction and add cache-busting for audio cues

devin-ai-integration[bot]
and
marksftw
committed
fix: rename audio cue files to mic-on-full.wav / mic-off-full.wav to bust mobile Safari cache

devin-ai-integration[bot]
and
marksftw
committed
fix: boost audio cue gain to 3x for iOS volume parity with TTS, remove debug buttons

devin-ai-integration[bot]
and
marksftw
committed
fix: remove extra blank line left by debug button removal (prettier)

devin-ai-integration[bot]
and
marksftw
committed
fix: prevent race condition in exitVoiceMode async recorder cleanup

devin-ai-integration[bot]
and
marksftw
committed
fix: increase audio cue gain to 7x, handle app backgrounding during voice loop

devin-ai-integration[bot]
and
marksftw
committed

Commits on Mar 7, 2026

fix: allow TTS to continue playing in background, only stop mic on backgrounding

devin-ai-integration[bot]
and
marksftw
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: implement voice mode improvements with continuous loop, audio error recovery, and TTS discovery#459

feat: implement voice mode improvements with continuous loop, audio error recovery, and TTS discovery#459
devin-ai-integration[bot] wants to merge 31 commits intomasterfrom
devin/1772755465-voice-mode-improvements

Commits on Mar 6, 2026

feat: implement voice mode improvements with continuous loop, audio error recovery, and TTS discovery

style: fix Prettier formatting in RecordingOverlay and UnifiedChat

fix: track recording start time with ref, fix stale closure in handleVoiceDiscard

fix: use startRecordingRef in voice continuation effect, fix || vs ?? for savedDuration

fix: add isTauri() guard for iOS TTS platform check, restart recording on empty blob

fix: use refs for exitVoiceMode recorder check and stale handleSendMessage/handleTTSDiscovery

style: fix Prettier formatting for ref type declarations

fix: call cancelTTSGeneration/stopTTS unconditionally in exitVoiceMode to avoid stale closure

fix: clear voiceState after non-voice retry success, exit voice mode on generation error

fix: re-throw TTS playback error so speakAndWait rejects and voice mode exits

fix: catch errors in speak() wrapper to prevent unhandled rejection in TTSButton

fix: reset voiceRetryCount and recordingDuration when dismissing error overlay via X button

fix: TestFlight bugs - show playing state in compact overlay, iOS audio cues, TTS auto-play after model download

style: run Prettier on RecordingOverlay and UnifiedChat

fix: reset timer on re-entering recording state, remove duplicate mic-on audio cue

fix: prevent AudioContext leak and handle empty preprocessed TTS text

fix: exit voice mode on startRecording failure, guard stale audioContextRef cleanup

fix: TestFlight round 2 - iOS audio cues resume(), duration flash fix, compact layout gap, TTS auto-play after download

debug: add test buttons for audio cue debugging on mobile Safari

fix: cancelGeneration now stops playback to prevent audio during async decoding window

fix: play audio cues before mic activation and after mic deactivation on iOS

fix: restore audioSession type after cue playback to allow mic capture

fix: use play-and-record audioSession type instead of playback

chore: replace audio cue WAV files with louder, trimmed versions

fix: remove 0.5 gain reduction and add cache-busting for audio cues

fix: rename audio cue files to mic-on-full.wav / mic-off-full.wav to bust mobile Safari cache

fix: boost audio cue gain to 3x for iOS volume parity with TTS, remove debug buttons

fix: remove extra blank line left by debug button removal (prettier)

fix: prevent race condition in exitVoiceMode async recorder cleanup

fix: increase audio cue gain to 7x, handle app backgrounding during voice loop

Commits on Mar 7, 2026

fix: allow TTS to continue playing in background, only stop mic on backgrounding