capcut-cli

English | 中文

Privacy — update to 0.18.0 if you have ever run capcut fixture. In every version up to and including 0.17.2, the bundle that command produces carries the device_id, mac_address and hard_disk_id CapCut stamps into your drafts. Only home paths and email addresses were redacted, and SANITIZE_REPORT.json wrote its own source_dir / out_dir unredacted, putting the username back. Because the documented flow is to attach that bundle to a public issue, following it published a stable device ID and MAC address while the filename and the report both said "sanitised". Fixed in 0.18.0 (#59). Treat any bundle generated by an earlier version as unsanitised. npm install -g capcut-cli@latest.

Security — update to 0.17.1 or newer. Versions up to and including 0.17.0 build the automation script behind export --batch by pasting the draft folder's name into it, so a folder named with the right characters can run commands of its own on macOS and Windows. Fixed in 0.17.1, together with an ffmpeg filter option injection reachable from a draft's caption colour (render --burn-captions), a compile spec whose name could write outside the draft store, predictable temp files on every draft write, and serve echoing credential values into its own output. Both injection paths need a draft folder or draft file you did not author, so the exposure is local rather than remote. npm install -g capcut-cli@latest. Details in the changelog.

Disclaimer: This is an independent, community-maintained project. It is not affiliated with, sponsored by, or endorsed by CapCut, JianYing, or ByteDance Ltd. "CapCut" and "JianYing" (剪映) are trademarks of ByteDance Ltd. All product names, logos, and brands are the property of their respective owners and are used here only for identification (nominative) purposes.

An independent CLI for CapCut / JianYing that any LLM agent can drive — zero dependencies, no server, both namespaces in one binary.

JSON in, JSON out: every command reads and writes the local draft store directly, with no MCP server or HTTP daemon. On newer CapCut versions it detects and synchronizes every readable timeline target instead of assuming draft_content.json is the only source of truth. That gives any model (Claude, DeepSeek, GLM, Kimi) a deterministic boundary for inspection, building, subtitles, captions, translation, and long-form cuts.

Use it three ways:

• CLI — npm install -g capcut-cli, then capcut <command> <project>
• Library — import { loadDraft, lintDraft, saveDraft } from "capcut-cli" (typed, zero-dep)
• Queue runner — capcut serve reads JSONL jobs from stdin, for n8n / Make / Coze

New in v0.19.1: every multi-range text highlight was being written past the end of the text it styled. styles[].range holds UTF-16 code units, not UTF-16LE bytes, so text-ranges, caption --karaoke, --highlight-words and any preset carrying text_ranges stored offsets twice as large as they should be — a plain add-text looked fine only because a full-span range clamps back to the end of the text (#85, measured by @hillimited across 38 app-authored drafts). Fixed everywhere those offsets are read or written, and lint --fix repairs drafts written by earlier versions (text-range-doubled). Full details in the changelog.

New in v0.19.0: a long render no longer fails with ERR_CHILD_PROCESS_STDIO_MAXBUFFER on a draft that was fine, and a render that does fail now names the missing decoder, encoder or filter instead of dumping ffmpeg's output; render --progress streams progress so a ten-minute job stops looking hung. lint gained three checks — caption reading speed (--max-cps), vertical safe-area (--safe-area) and speed consistency — and lint --fix can finally re-wrap CJK captions without disturbing per-character styling. Plus export-timeline no longer rounds a sub-half-frame clip to a zero-length OTIO clip (#82), and the keyframe docs now name the property_type the code actually writes (#80). No command was removed and no existing flag changed meaning. Full details in the changelog.

Install

Prerequisites: Node ≥ 18 (built-ins only — no native modules). Optional tools unlock specific commands: Whisper for caption, FFmpeg for render, ffprobe for automatic media metadata, and ANTHROPIC_API_KEY for translate.

npm install -g capcut-cli      # or: npx capcut-cli <command>

Build from source instead: git clone https://github.com/renezander030/capcut-cli && cd capcut-cli && npm install && npm run build (then npm link to expose capcut).

Quickstart

capcut doctor                                  # verify Node, FFmpeg, whisper, draft dirs
capcut quickstart my-first --video clip.mp4    # create + add input + lint, prints the "open in CapCut" step
capcut info ./my-first/                         # inspect the draft (add -H for a table)

Then open the project in CapCut to review and render. Every short-video platform forbids automated upload, so the publish click stays human.

Commands

JSON by default (pipe to jq); add -H for a human-readable table. Pass --jianying to use the JianYing enum namespace. Run capcut <command> --help for full flags.

Full reference for every command, option, and exit code: docs/command-reference.md (简体中文: docs/command-reference.zh-CN.md).

Sponsor

capcut-cli is MIT and free forever. Sponsoring funds faster releases and same-week support for new CapCut / JianYing versions — and unlocks power-user extras:

• $5/mo · Supporter — sponsors-only release notes plus your name in BACKERS.md. Keep the project moving.
• $25/mo · Pro — invite to the private capcut-cli-pro repo: premium template and caption-style packs, the full Claude viral-shorts pipeline, ready-to-run compile specs, and early-access builds. Plus priority issue triage.
• $100/mo · Team — everything in Pro for up to 5 teammates, written commercial-use confirmation, your logo in this README, and priority fast-tracking of the features your team needs.

Become a sponsor →

Using capcut-cli at work? The Team tier pays for itself the first afternoon it saves your engineers.

How it works

CapCut/JianYing store each project as local JSON. capcut-cli loads that store, validates against a version-aware schema, applies your edit, and writes it back atomically (with a .bak). No project files are uploaded anywhere; nothing runs as a service. See docs/version-support.md for the CapCut/JianYing versions and schema flags it understands.

Docs & examples

• docs/command-reference.md — every command and flag (简体中文)
• docs/quickstart.zh-CN.md — 剪映快速上手 (JianYing-first quickstart, Simplified Chinese)
• examples/ — end-to-end recipes (VO alignment, serve automation, batch subtitle correction)
• docs/version-support.md · docs/jianying-encryption.md
• CHANGELOG.md · Releases — what's new
• draftcat — sibling project: governed AI pipelines (Go, MIT), same single-binary, no-API design

Trademarks

CapCut™ and JianYing™ (剪映) are trademarks of ByteDance Ltd. This project is unofficial and is not affiliated with or endorsed by ByteDance; the marks are used nominatively to describe interoperability.

License

MIT