Skip to content
sysout.dev

In My Head 0.4 Release Notes

Transcription during recording, saved chunk progress, and recovery after interruptions.

Highlights

  • Start transcribing completed audio chunks while recording continues, reducing the remaining work after Stop. The microphone continues independently of transcription failures.

  • Save each successfully transcribed chunk. Reopen the app after an interruption to continue unfinished processing without repeating saved chunks. Failed work retains its audio and progress for Retry.

  • Keep automatic Model polish after recording stops and all transcription is complete. Recovery during polish resumes that stage without repeating Whisper.

  • Preserve existing recordings, transcript versions, edited prefixes, and session history through the database upgrade. Appending microphone audio creates a separate processing range and adds its text once.

  • Show active transcription with the existing rotating mark, coloured chunk bars, and animation, including when capture or the durable queue status differs. Numerical chunk counts remain in the notification only.

  • Retain CPU-based worker-pool selection. The separate OpenMP experiment is archived after device comparisons showed no benefit on the test phone.

Google Play "What’s new"

Transcription now starts while you record and saves progress as it goes. If processing is interrupted, reopen the app to continue from saved progress. Earlier transcript versions are preserved. Automatic Model polish starts after recording and transcription finish. Clearer activity feedback keeps ongoing work visible.

Release scope

Version code 4, version name 0.4. The manual CI release uses RELEASE_TAG=v0.4 and uploads the signed build and version-specific notes to Internal testing. Production promotion remains a separate release decision.

The proposed combined Recording/Transcribing layout is deferred. This release retains the current layout and its corrected live transcription feedback.