Install Jianying Professional 11.5.0 on an Apple Silicon Mac, write your edit as a structured JSON plan, then run the headless_draft.py build script with the path to that plan — Jianying Headless turns the plan into a native, fully editable Jianying draft, and can export a finished MP4 through Jianying's own engine without you ever opening the timeline. That is the whole workflow behind how to use Jianying Headless, and it is why the project drew attention when AlphaSignal covered it on 18 September 2026: it gives coding agents and scripts a structured way to drive a real video editor, producing drafts a human can still open, tweak and re-export in the normal Jianying interface.
🔥 Get the Agent OS as a free bonus: AI Profit Boardroom members get the full Agent OS zip, prompt libraries, daily tutorials and weekly live coaching calls. → Get inside · Want AI SEO help 1-on-1? Book a free SEO strategy session →
Everything here comes from the jianying-headless repository README as it stood on 21 September 2026, plus that AlphaSignal coverage. One thing to know before you invest an afternoon: the project describes itself as a private source preview with a personal, learning-focused, non-commercial licence — commercial use requires written authorisation from the author. Factor that in before building client work on it; more on the licence below.
What Jianying Headless Actually Does
Jianying is the Chinese sibling of CapCut, and Jianying Headless is a local automation toolkit for the macOS Professional edition. Its pipeline is simple to describe: materials plus an editing plan go in, an editable native draft comes out, and Jianying's own engine renders the export. Three capabilities make that useful, per the README. It builds native drafts from a JSON editing plan — multi-track video with speed and volume adjustments, locally imported video, images and audio, subtitles and text with local font embedding, and basic keyframe animation covering position, scale, rotation and opacity. It modifies existing multi-track projects inside isolated copies, so your original project is never overwritten. And it exports H.264/AAC MP4s through the genuine Jianying engine in an isolated process, started only on an explicit command.
The killer detail is the output format. Because the result is a real Jianying draft rather than a rendered file, automation handles the assembly grunt work while a human keeps full editorial control afterwards — open the draft, nudge the cut, fix a subtitle, re-export. That is a fundamentally different proposition from code-only rendering stacks like the one covered in the Claude Code Remotion guide, where the output is programmatic video with no editor round-trip.
If you want AI systems doing the boring 80% of your content production while you keep the creative control, check out the AI Profit Boardroom → get 1,000+ done-for-you AI workflows plus weekly live coaching. Want a personal plan for your content machine? Book a free SEO strategy session.
How To Use Jianying Headless: Step By Step
The requirements are strict, and the README is blunt that version matching is enforced — this is reverse-engineered tooling, so it targets exact versions rather than promising one-click compatibility with whatever you have installed:
- Check your hardware and OS. You need an Apple Silicon Mac running macOS 26.0 or later. Intel Macs and Windows are not supported.
- Install the exact Jianying version. The toolkit targets Jianying Professional 11.5.0, with 11.4.2 listed as compatible. Any other version is unsupported territory — the project explicitly does not provide lossless conversion across arbitrary Jianying versions.
- Install the supporting tools. Python 3.9 or later, FFmpeg with ffprobe, and the Xcode Command Line Tools.
- Prepare your materials and plan. Gather your clips, images, audio and fonts locally, then describe the edit in the plan's JSON format — tracks, clips with their timings, speed and volume settings, subtitle entries, and any keyframe animations.
- Build the draft. Run: python3 skills/yichen-jianying-edit/scripts/headless_draft.py build --plan /path/to/plan.json — the script validates the referenced resources and writes a native draft you can open in Jianying immediately.
- Export when ready. The local exporter runs as a separate, isolated process and only after an explicit command — nothing renders behind your back — producing a standard H.264/AAC MP4.
For agent-driven workflows the interesting step is number four: an LLM writes the JSON plan. The repository ships a standalone Agent Skill for exactly this, which is how coding agents such as Codex — and equally Claude Code — assemble editable video drafts from a text brief, per the AlphaSignal report. If you already run agents that follow project instructions, the pattern will feel familiar from the Claude Code AGENTS.md guide: define the rules once, let the agent produce structured output that a deterministic tool executes.
When something goes wrong, it is almost always one of two things. First, version drift: if Jianying auto-updated past 11.5.0, drafts may fail to open or build — the README enforces strict version and component matching for exactly this reason, so pin the app version and turn off automatic updates on the machine that runs the pipeline. Second, resource paths: the build step validates every referenced asset before writing a draft, so a plan pointing at a moved clip or a missing font fails early and loudly, which is far better than a silently broken timeline. Keep all assets in one project folder, reference them consistently, and re-run the build after any fix — drafts are cheap to regenerate because nothing renders until you explicitly ask for the export.
Where It Fits In A Content Automation Stack
Think of Jianying Headless as the assembly layer between your content pipeline and a human-quality edit. A realistic setup looks like this: an agent drafts the script and shot list, a generation tool produces or collects the clips — the making money with Sora guide covers the generation side — and Jianying Headless assembles everything into a draft that you or an editor polishes in fifteen minutes instead of building for two hours. The output stays native and editable, so quality control lives exactly where it should: at the end, with a human, on a real timeline.
That assembly niche has been the missing piece for faceless channels, clip repurposing and client content services. Rendering tools could already automate output, but they produced final files — any fix meant re-running the pipeline. A draft-first tool changes the economics of iteration, which is where most video time actually goes. Builders comparing approaches should look at the Hermes AI video generator guide for the fully generative route, and the free AI tools for content creators round-up for the supporting cast around whichever spine you choose. The orchestration principle is the same one behind Agent OS — agents plan, deterministic tools execute, humans approve — and if you are choosing which model writes your edit plans, the Goldie Bench write-up covers how the current agent models compare on structured tasks like this.
The Licence And The Fine Print
Read this section before you build anything that earns money. The README states the project uses a personal, learning-oriented, non-commercial licence, with commercial use requiring written authorisation, and third-party components keeping their original licences. It also describes itself as a private source preview, which means availability and terms could shift as the project evolves. On top of that, this is reverse-engineered tooling around a commercial application — Bytedance ships Jianying's own terms, and a Jianying update can break the toolkit until it catches up, which is exactly why the strict version pinning exists.
The practical read: treat it as a powerful lab tool for learning agent-driven video assembly, prototype your workflow with it, and get written authorisation — or wait for the licensing picture to mature — before it touches client deliverables. The skills you build transfer regardless: structured edit plans, agent-written JSON and draft-first pipelines are patterns, not products, and they will outlive any single tool. If you are earlier in that journey, start with the fundamentals in the how to build an AI agent guide and the best AI agents for coding round-up, then come back to video assembly once your agents reliably produce clean structured output.
Is It Worth Setting Up?
Yes, if you match the hardware profile and you produce enough video for assembly time to be a real cost — the requirements are narrow, but inside them this is the most editor-friendly automation route currently documented for the CapCut/Jianying family. No, if you need Windows, Intel hardware, or a clean commercial licence today. Either way it signals where content automation is heading in late 2026: agents writing structured plans, native tools executing them, and humans moving from operators to editors-in-chief.
If you want a content machine that runs while you sleep — built on workflows other members are already profiting from, check out the AI Profit Boardroom → join 3,000+ AI operators inside. Prefer to start with expert eyes on your plan? Book a free SEO strategy session — no cost, no catch.
