SuperSubtitler v0.3.0

SuperSubtitler 0.3.0 — who is speaking?

Download both files into one folder: SuperSubtitlerGUI.exe (the window) and SuperSubtitler.exe (the engine).

🗨️ Speaker changes in the subtitles

• New option “Mark speaker changes” (console: -speakers): the voices are told apart, and a new subtitle starts when the speaker changes.
• A quick exchange of two speakers shares one subtitle as - lines, the way professional subtitles show dialogue — only when both lines are complete and short enough, never a cut-off sentence.
• Runs on the processor while the graphics card already works on the next file. Each file shows how many speakers were found. If the recognition ever fails, you still get the subtitle, just without the marks.
• Know the cast? -speakers:N (or Number of speakers in Settings) fixes the count; How readily voices count as different tunes the automatic one.
• Uses sherpa-onnx 2.0.1 with the pyannote segmentation 3.0 and 3D-Speaker CAM++ models — pinned and SHA-256-checked like everything else, from the project's own mirror first. Downloaded once (about 55 MB) the first time you use it; nothing else is reinstalled.

🈶 Chinese and Japanese subtitles

• Words are no longer separated by spaces that do not belong there, lines break between characters and never start with punctuation, and the line length counts characters (it used to count bytes, so lines were cut far too short).

The exes are not code-signed; Windows SmartScreen may warn — "More info" → "Run anyway". Built by GitHub Actions.