jianying-editor-skill: The Jianying Editor Skill That Builds Your CapCut Drafts

What jianying-editor-skill actually solves

The repetitive part of short-video production is not the creative decision, it is the assembly. Dropping clips onto a track, splitting a voiceover into subtitle lines, matching each line to the right footage, picking a track from the Jianying music library. The README frames the project as a skill that takes a natural-language instruction and returns a finished draft: copywriting, voiceover, subtitles, music selection, effects and export. The intended user is someone who already works in JianyingPro, the Chinese desktop edition, and wants the timeline pre-built. It is not a renderer and not a CapCut tool. The README states plainly that it only targets the domestic JianyingPro desktop build and that CapCut International and mobile Jianying are unsupported, so anyone outside that audience is looking at the wrong repository. The scope is also narrower than the feature table suggests: the README says real-time effects such as smart cutout, beauty filters and speech-recognition subtitles cannot be called through code, and built-in Jianying AI features like one-click movie generation cannot be triggered. What remains is the mechanical assembly work, which is exactly where the time goes.

How the skill drives JianyingPro drafts

The mechanism is draft-file generation, not plugin hooks. The skill writes the `draft_info.json` structure that JianyingPro reads from its project directory, so the app sees a normal project when it next opens. The repository layout supports this: `scripts/vendor/` holds runtime dependencies such as `pyJianYingDraft`, `SKILL.md` is the instruction file handed to the AI editor, and `examples/` contains runnable scripts like `simple_clip_demo.py`, `my_first_vlog.py` and `compound_clip_demo.py`. Media handling is where the recent work went. Version 1.7, dated 2026-09-11, fixes the Jianying Pro 5.9+ "media lost, relink before editing" error by generating stable non-empty `local_material_id` values for `VideoMaterial` and `AudioMaterial`, and by staging external assets into the draft directory so cleanup of temporary files cannot corrupt the project. The same release removes a virtual-path fallback for cloud music and makes download failures raise an explicit error instead of producing a broken draft. On macOS, when `pymediainfo` or `libmediainfo` is missing, media parsing falls back to `ffprobe`, and `media_normalizer.py` regularizes non-standard video geometry that would otherwise crash Jianying's parser. That is a real architectural commitment: the skill is only as reliable as its guesses about a closed-source file format.

Installing the skill and cutting a first Vlog

The README recommends the Windows one-click script, which handles the code download, the directory layout and the Python libraries. It runs in PowerShell:

irm is.gd/rpb65M | iex

If you prefer to place the skill yourself, the clone target depends on your editor. Claude Code expects it under `.claude/skills/jianying-editor`, Trae under `.trae/skills/jianying-editor`, and Antigravity or Gemini Code Assist under `.agent/skills/jianying-editor`. A generic install into a root `skills/` directory is also documented:

# generic install

git clone https://github.com/luoluoluo22/jianying-editor-skill.git skills/jianying-editor

After cloning, the README asks you to install Python dependencies and initialize the browser capture environment used by the Web-to-Video feature:

pip install -r requirements.txt
playwright install chromium

The dependency list pins `uiautomation==2.0.20`, `playwright==1.52.0`, `edge-tts==7.2.3`, `pymediainfo==7.0.1` and `opencv-python==4.11.0.86`, among others. The skill probes for your Jianying install automatically; if that fails, you tell the AI the path directly, for example `C:\Users\Administrator\AppData\Local\JianyingPro\User Data\Projects\com.lveditor.draft` on Windows or `/Users/<name>/Movies/JianyingPro/User Data/Projects/com.lveditor.draft` on macOS. Then you speak to the editor in plain language. The README's example is: "Take the videos and photos in D:\旅行素材 and cut them into a Vlog, with light music and the title '周末露营记'." Expect no new draft to appear in Jianying's list until you restart the app or open and exit an old draft, because the README notes that Jianying does not refresh its file list live.

The export problem, and where this is the wrong tool

Automated export is the weakest part of the project, and the README says so. It works by simulating mouse and keyboard input through Windows UI Automation, which means you cannot touch the machine while it runs, and the automation is documented as steadiest on Jianying 5.9 or earlier. Newer builds can break it through changed dialogs or controls. On macOS there is no automated export at all; the README instructs Mac users to generate the timeline with the skill and finish the render by hand inside Jianying. That split matters for anyone planning unattended runs. A batch of twenty videos that must render overnight is not a supported workflow on Mac, and on Windows it depends on a pinned older Jianying version. There is a second boundary worth naming: the project cannot call Jianying's GPU-backed real-time features, so smart cutout, beauty filters and speech-recognition subtitles stay manual. If your pipeline needs a headless renderer with no GUI in the loop, this is the wrong layer entirely. The skill builds projects; Jianying renders them.

How it differs from using pyJianYingDraft directly

The obvious alternative is to write against `pyJianYingDraft` yourself, since the skill vendors it under `scripts/vendor/`. The difference is the layer above the library. With the raw library you construct tracks, segments and materials in Python, and you own every path decision, media probe and draft-directory guess. The skill adds the agent-facing instruction file, the natural-language entry point, the platform probing, the `ffprobe` fallback and the normalizer, plus the UI automation for export. If you are comfortable writing the draft JSON yourself, the library alone is less machinery and fewer pinned dependencies. If you want an editor agent to translate "make a 60-second commentary on D:\电影片段.mp4" into a timeline, the skill is the part that does the translating. The trade-off is transparency: with the skill you are trusting an LLM plus a set of heuristics to choose footage and structure, and the README's semantic matching feature, introduced in v1.4, is described as analyzing picture content, narration audio and SRT subtitles together to pick matching shots. That is a heuristic, and it will sometimes pick the wrong shot.

Maintenance, updates and the licence question

The repository is not archived and the last push was on 2026-09-11, which is recent. The changelog shows a steady cadence through 2026: v1.2 in January, v1.3 in February, v1.5 in April, v1.6 in June, v1.7 in September. Updating an installed skill is a `git pull` inside the skill directory, for example `cd .agent/skills/jianying-editor && git pull`, after which you may need to rerun `pip install -r requirements.txt` if dependencies changed. The cost to weigh is version coupling. Auto export is documented as steadiest on Jianying 5.9 or lower, so upgrading JianyingPro can break the export path, and each Jianying release can change the draft format the skill writes. The licence is reported as NOASSERTION, meaning GitHub could not match the `LICENSE` file to a known template; read that file yourself before shipping the skill inside a commercial product, and note that `scripts/vendor/` bundles third-party code with its own terms. Nothing here is legal advice.

What the documentation does not tell you

Several things a prospective user would want are absent. The README does not document rollback if a generated draft corrupts an existing project, beyond the auto-healing mechanism mentioned in v1.5 that detects and repairs damaged or outdated project files. There is no published benchmark for how long a draft takes to generate at a given clip count, and no stated limit on project size. The export automation's failure modes are described as sensitivity to dialogs and controls, but not enumerated. The FAQ on the project homepage covers the draft-refresh and export questions the README already answers. For a tool that writes into a closed-source application's project directory, the absence of a documented draft-format version matrix is the gap that will bite first: you find out a format changed when Jianying refuses to open the project.

Editorial conclusion

Adopt it if you already edit in JianyingPro on Windows or macOS and want the timeline, subtitles, voiceover and music assembled for you before you open the app. Skip it if you work in CapCut International or on mobile, or if a headless render pipeline is your requirement, because the project leaves final rendering to Jianying itself. Before installing, confirm two things: which draft directory your Jianying build actually uses, and which Jianying version you have, since auto export is documented as steadiest at 5.9 or below.

Frequently asked questions

Official sources

1. Issues
2. luoluoluo22/jianying-editor-skill on GitHub
3. Project website
4. README