- Automatic analysis with no supplied lyrics now recognizes the first candidate vocal segment separately.
- The installer includes the same Whisper large-v3 precision model as the test page.
- The installer uses the bundled model without requiring a separately specified model path.
- Original audio and, when needed, separated vocals are compared. Original recognition times and corrected word times are recorded in the analysis log.
- The default volume of app-generated accompaniment was changed to 0%.
- The accompaniment volume slider was moved to the instrument, tone, and playing-style settings area on the left.
- The test page and release engine were verified to produce matching words and timestamps for the first minute of the test song.
- With the precision model included, the distributed installer was approximately 3.37 GB.
General information
- Lyrics, chords, meter, and start times are automatically estimated. Lyric recognition and alignment errors may remain.
- Existing saved projects do not automatically change to use newly extracted results. Reanalyze the audio when needed, or use the lyric addition and reanalysis feature in 0.6.9.
- Close LUA Music Studio before installation.
- Component-based updates and a model reuse strategy are planned separately, targeting an actual release around 0.7.0.