CapCut SmartCut

You edit CapCut projects programmatically via the bundled Python scripts in scripts/. Everything runs locally against the CapCut drafts folder — no MCP server, no API keys required.

Setup (first use only)

The scripts need a venv with pydantic. Check it exists, create if not:

[ -x ~/.claude/skills/capcut-smartcut/scripts/.venv/bin/python ] || bash ~/.claude/skills/capcut-smartcut/scripts/setup.sh

Add --with-openai to setup.sh only if the user wants GPT-enhanced duplicate detection (also needs OPENAI_API_KEY in the environment).

Running commands

Everything goes through one CLI (JSON in, JSON out):

~/.claude/skills/capcut-smartcut/scripts/.venv/bin/python \
  ~/.claude/skills/capcut-smartcut/scripts/cli.py <command> --args '<json>'

Command	Purpose	Key --args fields
list	List all CapCut projects	drafts_dir (optional, auto-detected)
open	Inspect a project: video segments, subtitles	project_name (partial match) or project_path
smart-cut	Cut silences + duplicate takes (heuristic first pass)	project_name, silence_threshold_sec (default 1.0), similarity_threshold (0–1, default 0.5), use_openai (default false)
cut-ranges	Apply exact cuts YOU decided (duplicates, filler)	project_name, ranges (array of {start_sec, end_sec} on the current timeline; auto-merged)
edit-subtitle	Change a subtitle's text/timing	project_name, segment_id, new_text, new_start_sec, new_duration_sec
split-subtitle	Split a subtitle at a timeline point	project_name, segment_id, split_time_sec
merge-subtitles	Merge two subtitles (A absorbs B)	project_name, segment_id_a, segment_id_b
fix-word-timing	Set word-level timing arrays	project_name, segment_id, words_text, words_start_ms, words_end_ms (all same length, ms relative to segment start)
batch-edit	Multiple subtitle edits in one pass	project_name, edits (array of {segment_id, new_text?, new_start_sec?, new_duration_sec?})

Example:

.../cli.py smart-cut --args '{"project_name": "Podcast Episode 5", "silence_threshold_sec": 1.0}'