Natural-language Jianying editing skill

Document type
Field manual
Updated
Reading time
9 min

jianying-editor-skill lets Claude Code, Cursor, Trae and Antigravity automate Jianying Pro scripts, voiceovers, subtitles, music, effects and MP4 exports.

SourceOpen original report
Document details
luoluoluo22/jianying-editor-skill

jianying-editor-skill is an open-source Agent skill that lets Claude Code, Cursor, Trae and Antigravity operate Jianying Pro through natural language, automating scripts, voiceovers, subtitles, music, effects, screen recording and MP4 export. Python scripts drive editing actions; the native Jianying application still renders the final output.

GitHub Stars
2.2k
Automation capabilities
11
Maximum export quality
4K
Open-source license
MIT

01Tool overview

Agent Jianying operation skill

jianying-editor-skill is a set of skill files installed in an AI editor. Claude Code, Cursor, Trae and Antigravity read its rules and convert natural-language editing requests into Jianying Pro operations. Ask it to turn a folder of videos into a Vlog with upbeat music and narrated subtitles; it handles the script, voiceover, subtitles, music and MP4 export.

It consists of Python scripts and Markdown rules, not a new video editor. Scripts directly read and write Jianying drafts through bundled pyJianYingDraft, while Markdown rules in rules/ tell the Agent which workflow to use for a task. The native Jianying application still renders the final output; the skill operates Jianying and does not replace it.

v1.5 added automation: automatic repair detects and repairs damaged or outdated drafts; cloud music library extraction retrieves previously used Jianying cloud music IDs for AI reuse; semantic asset matching analyzes narration and selects matching footage. The workflow has five fixed stages, with each stage automatically consuming the previous stage's output.

  1. Initialize

  2. Route request

  3. Execute operations

  4. Integrate assets

  5. Export

Describe the request in natural language; AI handles the script, voiceover, subtitles, music and export in one workflow.

— jianying-editor-skill README · One-sentence definition · Translated from Chinese

02Installation

Installation for four Agents

Prepare Jianying Pro 5.9 or earlier, Python and Git. Windows users can install with one PowerShell command; on other platforms, manually clone to the path for your AI editor.

One-command Windows installation

bash
# Windows · One-command installation(PowerShell)
irm is.gd/rpb65M | iex

Manual installation paths by editor

Each AI editor uses a different skill directory. Clone the repo to the corresponding location; the commands differ only in their final path.

bash
# Antigravity / Gemini
git clone https://github.com/luoluoluo22/jianying-editor-skill.git .agent/skills/jianying-editor

# Trae IDE
git clone https://github.com/luoluoluo22/jianying-editor-skill.git .trae/skills/jianying-editor

# Claude Code
git clone https://github.com/luoluoluo22/jianying-editor-skill.git .claude/skills/jianying-editor

# Cursor / VSCode / General
git clone https://github.com/luoluoluo22/jianying-editor-skill.git skills/jianying-editor

Dependency installation

After cloning, install Python dependencies and download Chromium through Playwright for web-to-video recording and some automation tasks. Update later by entering the directory and running git pull.

bash
# Install dependencies
pip install -r requirements.txt
playwright install chromium

# Update later
cd .agent/skills/jianying-editor && git pull

03Capability reference

Eleven editing capabilities and scripts

Each capability has an independently runnable Python script, selected and chained by the Agent according to the task. The eleven capabilities are grouped by stage: assets handle imports and effects; audio handles voiceovers, subtitles and music; generation handles recording and film commentary; output handles export, animation and compound clips. Describe your request in natural language without memorizing script names; the Agent routes it automatically.

Assets · 01

add_media_safe()

Asset import

Automatically place videos, audio and images on the timeline, handling paths and track alignment.

Assets · 02

asset_search.py

Effects / Transitions / Filters

Search built-in assets by keyword and apply effects, transitions and filters without memorizing names.

Audio · 03

TTS · 原生 / 微軟

AI voiceover

Convert text to speech with native Jianying voices and Microsoft TTS across multiple languages and voices.

Audio · 04

add_narrated_subtitles()

Subtitle generation

Align subtitles automatically to the narration timeline; entrance and animation effects can be applied.

Audio · 05

build_cloud_music_library.py

Automatic music

Select BGM from local or cloud music libraries and recommend background music for the topic.

Generation · 06

web_recorder.py

Web-to-video

Record HTML / Canvas animations with Playwright to turn web pages into usable footage.

Generation · 07

recorder.py

Screen recording + smart zoom

Record the screen, automatically zoom toward the cursor and add red-circle markers for operation tutorials.

Generation · 08

movie_commentary_builder.py

Film commentary

AI analyzes video content and generates storyboards and voiceovers for film commentary.

Output · 09

auto_exporter.py

Automatic export

Export MP4 from 1080P to 4K with configurable resolution and frame rate.

Output · 10

jy_wrapper.py apply-zoom

Keyframe animation

Animate zoom, position and opacity through keyframes with CV-assisted smart zoom.

Output · 11

jy_wrapper.py clone

Compound clips

Nest multiple projects into a compound clip or copy a template into a new project.

Task and route reference

The Agent matches task types to rule files in rules/, then calls the corresponding scripts. These are the main routes recorded in SKILL.md.

TaskRouting rule filesCore script
Cloud video + BGM + voiceovermedia.md + audio-voice.mdcloud_video_music_tts_demo.py
Script to video (narration + subtitles)text.md + audio-voice.mdadd_narrated_subtitles()
Screen recording + zoomrecording.mdrecorder.py
Batch / headless exportcore.md + cli.mdrobust_auto_export.py
Film commentarygenerative.mdmovie_commentary_builder.py

04Core scripts · Official principles

Editing command-line tools

Natural-language instructions handle most tasks, but understanding the scripts helps with precise debugging and adjustments. These tools and mechanisms are documented in SKILL.md and README for direct terminal use. <SKILL_ROOT> in commands is your installation path.

Draft inspection

draft_inspector.py lists and inspects drafts. list --limit 20 lists recent drafts, summary --name "DraftName" shows a summary, and show --kind content --json exports the full structure.

Source · SKILL.md

Find assets by name

asset_search.py "關鍵字" -c filters searches built-in filters, transitions and effects by keyword; sync_jy_assets.py synchronizes the local Jianying asset index.

Source · SKILL.md

Automatic draft repair

During initialization, JyProject is instantiated with overwrite=True to detect and repair damaged or outdated drafts using the 5.9+ draft_info.json structure, preventing incomplete drafts from interrupting the workflow.

Source · SKILL.md · v1.5

Cloud music library extraction

build_cloud_music_library.py extracts previously used Jianying cloud music IDs into cloud_music_library.csv for AI reuse of familiar BGM. Text styles have a corresponding build_cloud_text_styles_library.py.

Source · README · v1.5

Semantic asset matching

AI analyzes narration or script semantics to select matching footage from your assets, reducing manual track alignment. v1.5 describes this as understanding content.

Source · README · v1.5

Headless batch export

auto_exporter.py "DraftName" "output.mp4" --res 1080 --fps 60 exports a specified draft without opening the interface for batch processing. Do not use the mouse or keyboard during export.

Source · SKILL.md

Web pages as footage

web_recorder.py --url "http://localhost:3000" --duration 5 uses Playwright to record HTML / Canvas animations, turning web pages into video footage for the timeline.

Source · SKILL.md

Environment checks

Run api_validator.py first to check API and environment readiness, then use wrapper commands such as jy_wrapper.py export-srt --name "DraftName" for individual operations.

Source · SKILL.md

05Usage example

Natural-language editing and MP4 export

This illustrative workflow gives the Agent a folder and a request, then routes through rules, inspects drafts, searches filters, generates narration and subtitles, adds cloud music and exports headlessly at 1080P. Commands use actual SKILL.md syntax; asset search results are illustrative and depend on your Jianying asset library.

claude code · jianying-editor-skill v1.5


$ You ›
  把 D:\trip 的影片和照片剪成 Vlog,配輕快音樂、
  加旁白和字幕,最後匯出 1080P。


# [loaded .claude/skills/jianying-editor/SKILL.md]
# [route → rules/media.md + rules/audio-voice.md]
# [init JyProject(overwrite=True) · 自動修復檢查通過]


claude: Claude ›
  先盤點現有草稿與素材,再開始組裝時間軸。


> python scripts/draft_inspector.py list --limit 20
  # 找到 8 個草稿 · 無同名衝突 → 建立 "Trip-Vlog"


> python scripts/sync_jy_assets.py
  # 已同步本地素材索引


> python scripts/asset_search.py "輕快" -c filters
  # 回傳符合關鍵字的濾鏡 / 轉場候選(示意)


# [匯入 D:\trip · 影片 x12 · 照片 x9 → 時間軸]
# [add_narrated_subtitles() · 生成旁白 + 對齊字幕]
# [查詢 cloud_music_library.csv · 依「輕快」推薦 BGM]


claude: Claude ›
  草稿已組好。剪映草稿不會即時刷新,稍後請切換專案查看。
  接著做無頭匯出——匯出期間請勿操作滑鼠鍵盤。


> python scripts/auto_exporter.py "Trip-Vlog" "D:\out\trip.mp4" --res 1080 --fps 60


  ok: 匯出完成 · D:\out\trip.mp4 · 1080P / 60fps
  ok: 時長 01:12 · 旁白 + 字幕 + BGM 已就緒

        

Provide a folder and a one-sentence request, and it returns an MP4 ready to publish.

— Natural language → Routing → Scripts → Export · Translated from Chinese

Routing and draft inspection

The Agent first matches "Vlog + music + narration" to rules/media.md and rules/audio-voice.md, then calls their scripts. Each step writes its output, including drafts, subtitles and music, into the Jianying draft for the next step to consume.

Use draft_inspector.py to inspect draft structure at any stage and intervene when needed. Jianying still renders the final output, with the same quality as manual operation.

06Boundaries and limitations

Platform and automation limits

07Next steps

Customization of routes and asset libraries

Routing logic is in Markdown rule files under rules/, and scripts are in scripts/. Both are readable and editable; adjust workflows and build asset libraries to match your editing habits.

Advanced uses

1. Customize routing rules. Open files such as rules/media.md and rules/audio-voice.md to change task workflows and defaults without modifying Python.

2. Build cloud asset libraries. Run build_cloud_music_library.py and build_cloud_text_styles_library.py to extract your usual cloud music and text-style IDs into a database for AI reuse.

3. Copy projects from templates. jy_wrapper.py clone --template "TemplateName" --name "NewProject" copies an existing project into a new draft, keeping a consistent style across a video series.

4. Create intros from web animations. Record your HTML / Canvas animations, such as a star-particle intro, with web_recorder.py, then import them to the timeline for a custom opening.

5. Create film commentary. Use movie_commentary_builder.py --video "video.mp4" --json "storyboard.json" for AI video analysis, storyboards and voiceovers to produce long-video commentary quickly.

Three further documents

① README.md: Installation, platform support, capabilities and limitations. ② SKILL.md: The Agent's specification, including script commands, execution loop and routes. ③ usage.md: More complete scenarios and examples.

This does not replace Jianying; it uses the native application to render and export.

— jianying-editor-skill README · Limitations · Translated from Chinese