PAI-Pro Filmmaking Studio
Turn scripts, visuals, video, and voice into a filmmaking workflow with a coding agent and local project files.
- Source repo
- Utopai-Research/pai-code
- Stars
- ★ 346
- Last updated
- 19d ago
- License
- NOASSERTION
- Primary language
- JavaScript
- FA score
- 45/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changesCodex · Claude Code
- Cost
- Free tier plus a paid hosted plan
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- An independent filmmaker uses Codex to configure a project, plan scenes on the canvas, and organize shot assets.
- Not a fit if
- Teams unwilling to use Docker or local startup scripts
- Teams needing image, video, and voice generation at no cost
- Source review
- 45/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
PAI-Pro is a local-first AI filmmaking workspace built around Claude Code or Codex, filmmaking skills, a visual canvas, and a timeline. Its agents can work with local project files and persistent context, using skills for scripts, images, video, and voice. The canvas supports scene design, asset management, and shot arrangement, while the timeline helps organize larger productions. One API service handles image, video, voice, and 4K upscaling; the README also documents BYOK. The project runs through Docker or host scripts and provides a local web interface, with a hosted web service also linked.
After installing and configuring PAI-Pro in Claude Code or Codex, the agent uses local project files, persistent context, and filmmaking skills for script, image, video, and voice tasks. Users can design scenes, manage assets, and arrange shots on the canvas, then organize a production with the timeline. Media operations are provided through the API service: generate_image, generate_image_pro, generate_video, generate_voice, and upscaler. The README gives quality, estimated time, reference limits, and estimated prices for each operation. The Docker deployment opens at http://localhost:7588; host mode uses http://localhost:7443.
- An independent filmmaker uses Codex to configure a project, plan scenes on the canvas, and organize shot assets.
- A screenwriter using Claude Code wants an agent to work on scripts and related production tasks through filmmaking skills.
- A small production team needs one workspace to manage scenes, images, video, and voice assets.
- A creator needs image generation at different resolutions with support for reference images.
- A team needs video or voice generation and a way to upscale finished video to 4K.
- A team wants to run the filmmaking interface locally through Docker.
How do you install or deploy this agent?
The README's quick start recommends asking Codex or Claude Code to read the setup documentation and install PAI-Pro; it says to provide a PAI_KEY and defaults to Docker. To start directly with Docker:
./scripts/docker-start.shFor Claude Code with Docker:
PAI_DEFAULT_AGENT_ID=claude ./scripts/docker-start.shHost mode uses a setup script followed by the startup script:
./scripts/setup
./scripts/start.shFor Claude Code in host mode:
./scripts/setup --agent claude
PAI_DEFAULT_AGENT_ID=claude ./scripts/start.shThe Docker interface is at http://localhost:7588; host mode is at http://localhost:7443.
How do you use this agent?
Start the app using one of the installation paths and configure PAI_KEY. The README also supports BYOK; provider setup and exact payloads are covered in the API Service documentation. Open http://localhost:7588 for Docker or http://localhost:7443 for host mode, then choose the agent from the terminal panel's header control. Use the agent for filmmaking skill tasks and organize scenes, assets, and shots in the canvas and timeline. Media generation is available through the API service operations generate_image, generate_image_pro, generate_video, generate_voice, and upscaler.
What are this agent's strengths and limitations?
- Explicitly supports Codex and Claude Code, with agent switching from the project terminal panel.
- Combines local project files and persistent context with filmmaking-specific skills.
- The canvas and timeline cover scene, asset, and shot organization for larger productions.
- One API service covers images, video, voice, and 4K upscaling, with estimated prices and times documented.
- Offers both Docker and host-script deployment paths.
- Media generation requires PAI_KEY; the README lists generation and upscaling as paid operations with estimated prices.
- Setup involves scripts and agent configuration; host mode requires separate setup and startup steps.
- The README does not specify supported operating systems, minimum hardware, or offline mode support.
- Commercial use is constrained by the PAI PRO Sustainable Use License; some skills or enterprise-designated source require a separate agreement.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| PAI-Pro Filmmaking Studio This agent | 45 · Major gaps | Self-hosted serviceFreemium | ★ 346 | 19d ago | JavaScript | Codex · Claude Code |
| Generative Media Skills | 41 · Major gaps | CLIFree + model costs | ★ 5.9k | today | Python | Claude Code · Claude.ai |
| ArcReel Video Creation Workspace | 52 · Major gaps | Self-hosted serviceFree + model costs | ★ 5.6k | 1d ago | Python | OpenAI API · Claude API |
| ReelMimic Style-Mimicking Video Studio | 50 · Major gaps | — | ★ 3.6k | today | JavaScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
The README calls the product local-first and lists Claude Code/Codex plus Docker and host modes. Its quick start asks the user for a PAI_KEY and says paid generations are staged first, giving some evidence about permissions and external effects (1). The supplied material does not explain key storage, media flows, permission scope, confirmation steps, or recovery, so those areas lose points; no rollback evidence is supplied (0). The license names a licensor and requires attribution (2), while publisher identity remains unknown as specified in the prompt.
The CI file specifies Node 20, lockfile-based installation, and automated server and web checks. This supports dependency availability (2), but configuration alone does not establish execution results. The README refers to the repository as pai-pro while the object is pai-code, and no implementation files corroborate the product description, so self-consistency is only partly supported (1). No failure-message or error-handling evidence is supplied (0).
The README identifies AI filmmaking as the audience and lists scripting, images, video, voice, canvas, and timeline scenarios. It also offers Claude Code/Codex and Docker/host choices (2). Capability limits and boundaries are not specific (1); the install prompt is broad and does not say which requests trigger which skills (1). It does name environment options and ports (2).
The README organizes quick start, API, resources, and skill documentation, and gives install commands and ports (2 each for information architecture and install notes). It links a FAQ whose contents were not supplied, and gives no concrete workflow example or detailed limitations (1 each). The object’s license metadata is NOASSERTION, but the repository includes a clear, restrictive custom license (2). No version history or changelog is supplied (0). Security reports have a private channel, and maintainers say they will respond when able, but no responsible maintainer or response target is identified (1).
The README describes an agent-driven filmmaking workspace and lists image, video, voice, canvas, and timeline workflows. Its API table gives capabilities, estimated times, and prices, allowing an initial assessment of use and cost (2 each). These are product claims without implementation or result evidence, so they do not merit full marks.
The README links its claims to specific documentation and CLI files, and CI lists check commands, giving claims some traceability (2). The cited documentation and implementation were not supplied, limiting corroboration across sources (1). Times, quality, and prices are described as estimates or promotional claims, without their basis, so fact and inference are only partly distinguished (1).
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- The supplied material omits implementation code and the setup, API, skills, architecture, and FAQ documents linked from the README; specific behavior, permission boundaries, and key handling cannot be verified.
- The object is identified as pai-code, while the README points to pai-pro; verify that this revision corresponds to the product under review.