Pot să vibecodez CapCut?

Build me a video captioning and clipping tool to replace CapCut Pro.
Requirements:

- A CLI: Python + ffmpeg + faster-whisper. clip input.mp4 transcribes and
  burns captions; --cut 0:12-0:58 trims first; a folder as input batches
  every video in it.
- Captions from word-level timestamps, grouped 3-5 words per line, rendered
  as ASS subtitles: bold font, white fill, black outline, centered in the
  lower third, burned in with ffmpeg.
- style.json controls the look in one place: font, size, colors, margin.
- Aspect exports: --format 9x16 (blur-pad or center-crop, chosen by flag),
  1x1, and 16x9, with bitrate defaults sane for phone playback.
- --srt writes the plain subtitle file next to the video so I can fix
  typos, then --from-srt re-burns without re-transcribing.
- Output is <name>-<format>.mp4 beside the input; never overwrite a source
  file.
- Fully local and offline after the model download: no accounts, no uploads,
  no telemetry. Use the GPU when present, CPU still works.
- Out of scope: a timeline editor, effects and templates, and a music
  library; CapCut's editor polish is the product. This is for talking-head
  clips.
- README: installing ffmpeg and choosing a whisper model size.