[Bug]: iOS voice input drops opening words when speaking immediately after tapping the microphone #17888

Before submitting

• I searched existing issues and did not find a duplicate.
• I included enough detail to reproduce or investigate the problem.

Area

apps/mobile

Steps to reproduce

1. Open a thread’s composer in the iOS app.
2. Tap the microphone and immediately say: “One, two, three, four, five. Please help me review this code and find the bug.”
3. Finish dictation and inspect the inserted transcript.
4. Repeat with the same sentence, waiting until recording has started before speaking.

Expected behavior

Voice input should preserve the beginning of speech when speaking immediately after tapping the microphone. Capture should start as soon as microphone permission and audio setup allow, even if the transcription model is still preparing.

Actual behavior

The opening words are consistently missing when I start speaking immediately. Waiting before speaking preserves them.

I tested the same sentence four times:

This happens on every attempt when I speak immediately and significantly disrupts my workflow. I habitually begin speaking as soon as I tap the microphone, as I do with other dictation tools.

Impact

Major degradation or frequent failure.

Version or commit

T3 Code iOS 2.0.0, latest available TestFlight build as of October 10, 2026. Exact build number not recorded.

Environment

iPhone 17 Pro Max running iOS 27.0.

Logs or stack traces

None collected.

Screenshots, recordings, or supporting files

The four manually collected transcription results are included above. No screen recording was collected.

Workaround

Wait until recording has started before speaking.

Possible fix

Start capturing audio while the on-device transcription model prepares, then transcribe the complete recording once preparation finishes. If preparation fails, stop recording, delete the temporary audio, and show an error.

Source inspection found that the current controller awaits transcription preparation before starting the recorder. This is consistent with the observed delay, although no device timing instrumentation was performed.