← All updates
v0.8.2Release notes

Better timing. Cleaner edits.

Fixes for audio that shifts or ends early, headings that wrap even when they fit, and exports that fail after many short scenes.

  • 01

    Audio timing fixes

  • 02

    Headings that fit

  • 03

    Undo canvas changes

Keep audio in sync through the final export

  • Mixing or normalizing a video's audio could shift the sound or cut it off before the picture ended. Those steps now preserve the source audio's timing and ending.
  • Using audio from another recording could fail when an edit kept several separate sections of that recording. Those audio sections now join in the requested order when you preview or export.

Fix headings that wrap too soon and previews that miss text

  • A short heading could wrap onto a second line even when it fit, while wider letters could run past the allowed width. Moviestar now measures the chosen font before deciding where to break each line.
  • Preset headings and labels could look oversized on a small video. Their default text size, spacing, and outlines now scale to the video canvas. Sizes you set explicitly in pixels stay as you specified them.
  • A screenshot of a single-video project could leave out the titles and captions you added. Those previews now include the stored text, even when you haven't created any scenes.
  • Choosing a transparent background for a title could make the render fail. That setting now renders as a transparent background.

Export short scenes and undo portrait/landscape changes

  • An edit with many very short scenes could gain extra time at each cut and eventually fail to export. Fixed-frame-rate exports now keep the scenes where you placed them. The final video can still end less than one frame later than the exact edit duration.
  • Switching an existing scene edit between landscape and portrait could fail once it had edit history. You can now make that change, then undo it while keeping earlier edits available.
  • When checking whether a camera crop stays steady between two moments, Moviestar could accidentally compare later frames too. The warning now checks the two requested frames, reducing false alarms and missed changes. You still need to review the shot.

Load quiet footage offline and finish transcription cleanly

  • Footage with an audio track but no detected speech could refuse to load offline when no Whisper model was cached. With local transcription installed, load --no-download can now finish with an empty transcript. Footage containing speech still needs a cached model.
  • Transcription could print a successful result and then hang instead of returning control to your agent. With the default settings, the command now exits after transcription finishes.

Get v0.8.2

Existing projects open without a migration. Recheck headings and captions after upgrading: measuring the selected font can change line breaks, and unsupported fonts still use an estimated fallback. The update installs an additional library for font measurement. Use the transcription extra below to turn speech into captions; supplied caption files and --no-transcribe still work with the base package. Local editing requires Python 3.10 or newer, FFmpeg, and ffprobe.

python -m pip install --upgrade 'moviestar[transcribe]==0.8.2'
moviestar doctor
Full technical release notes on GitHub ↗
← Previous release: v0.8.1 · More reliable edits. A clearer review.