PAI-Pro workflow
A local AI filmmaking studio: Claude Code or Codex writes the script and makes frames, voices and clips on a canvas
- 352 ★ on the repository
- Last commit
Broad access
Why this level
- Paid generations spend the balance of your PAI Pro platform account
- The app runs an unattended agent, and Docker mode keeps it inside a container
- Setup scripts put skills into the agent's directory, and the app opens a local web interface
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add workflows/pai-proDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Paid generations spend the balance of your PAI Pro platform account
- The app runs an unattended agent, and Docker mode keeps it inside a container
- Setup scripts put skills into the agent's directory, and the app opens a local web interface
This is third-party code. Review the repository files before installing.
What it does
PAI-Pro runs on your machine and connects your Claude Code or Codex to a set of filmmaking skills and a visual canvas with a timeline. Images, video, voice and 4K upscaling come from one paid service on the PAI Pro platform. The agent works in a terminal inside the app, and results land on the canvas as nodes with links, so you can see where each came from. The skills cover the stages of work. `script-compose` writes a screenplay and splits it into shots of up to 15 seconds. `image-compose` designs characters and storyboards, and `video-compose` animates frames and continues clips. `voice-compose` gives a character a voice. `groups-compose` gathers related nodes into labelled scene frames. `story-to-video-workflow` sets the order of work from script and references to clips on the timeline, and the Timeline tab plays finished clips back to back. One platform key replaces separate provider keys for each step. Paid generations are queued first. The app runs in Docker (the recommended path) or locally, and keys and balance are managed on the platform site.
Who it is for. For video makers, screenwriters and designers who assemble short AI films and want to drive the whole job from Claude Code or Codex.
Good fit when
- You have an idea for a short and want to go from script to clips in one project
- You need a recurring character: a look sheet first, then a voice and scenes with them
- A script has to be split into shots of 15 seconds and voiced
- A clip must be continued or restyled without losing its link to the source frame
- You want to see on a canvas which references produced each frame and play the result on a timeline
Not a fit when
- You need a commercial studio: the Sustainable Use license permits personal use, non-commercial research and a company's internal use
- You cannot pay for generations: images, video, voice and upscaling run through a paid service
- You need to edit footage you shot: the app assembles clips created inside the project
- Docker is not available to you: the project has a hosted web service, but that is a different route
Example request
Design a weathered detective in his forties in noir style, give him a gravelly voice and make a six-second clip of him standing in the rain outside a dinerLimitations
Generations need a `PAI_KEY` from the PAI Pro developer platform, which goes into the `.env` file. Generations are paid, and the platform shows prices and balance. The PAI PRO Sustainable Use License permits personal use, non-commercial research and a business's internal use; commercial work falls outside it.
New projects start with Codex by default. For Claude Code, set `PAI_DEFAULT_AGENT_ID=claude` before launch. The agents run generations differently: Claude Code makes background calls, while Codex waits for the result.
Whether the PAI Pro platform is reachable from Russia has not been checked. The repository is named pai-code, while the documentation calls the project pai-pro.
How to disable. Stop the app with `./scripts/stop.sh`, then delete the clone folder. In Docker mode projects and agent sign-in data live in named Docker volumes, which you remove with docker commands. Revoke the key on the developer platform.
FAQ
How do I run PAI-Pro with Claude Code?
Clone the repository, copy `.env.example` to `.env` and add your platform key. Then run `PAI_DEFAULT_AGENT_ID=claude ./scripts/docker-start.sh` and open http://localhost:7588. You can switch the agent later in the terminal panel header.
What does the PAI_KEY do?
The key gives a project one access to image, video and voice generation and to 4K upscaling. Keys, tasks, balance and credits are managed on the PAI Pro developer platform.
How is Docker mode different from local mode?
Docker is recommended for trying the app and for daily work: the image includes ffmpeg, poppler, cloudflared, Claude Code and Codex, and the unattended agent stays inside the container. Local mode is for people changing the server and web source, and it opens on port 7443.
Are there skills beyond the six main ones?
Yes, the Claude Code marketplace manifest also registers `add-note` and `show-dag`. They are small helpers inside the project agent template: the first takes notes, the second gives an overview of the node graph.
Can I try PAI-Pro without installing it?
Yes, the project has a hosted web version. The local studio stays the main route: it works with your own Claude Code or Codex and keeps projects on your machine.
Discussion
No questions or reviews yet.
Related
WorkflowsSkills, a CLI and an MCP server that let Claude Code or Codex make video from a brief: editing, graphics, voiceover, subtitles
A local video studio for your coding agent: describe a video in one sentence, the agent directs, your machine renders with no cloud
A skill for Cursor, Claude Code and Codex: from a product name or URL it builds a 4K promo with music, voice-over and checks
A video editing skill for Claude Code and Codex: the agent cuts pauses and filler words, grades color, burns subtitles and adds animated overlays
A skill that writes prompts for video, image and music generation on 14 platforms and can submit them in the browser
Two skills for generative video and images: video writes Seedance, Kling and Veo prompts by dramaturgy rules, image picks the model and its syntax
README in short
The quick start is a ready sentence you paste into Claude Code or Codex, and the agent installs the project itself. Below are a table of service capabilities with estimated prices and generation times, a resources list and the license terms.