ARIS-Movie-Director skill
A skill set that turns a fuzzy idea into a frame-based comic movie: script, storyboard, generation and cross-model review.
- 77 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The pipeline runs external CLIs and spends image provider credits
Install
- Works in
Manual install
python3 -m pip install -r requirements.txtRun in the repository clone. Then `python3 cli/preflight.py`. There is no installer.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The pipeline runs external CLIs and spends image provider credits
This is third-party code. Review the repository files before installing.
What it does
The repository turns a fuzzy story into a movie of images: the reference work has 19 scenes and 24 frames. The `comic-author` skill builds `comic.json` with locked identities and expected literals, `movie-pipeline` drives the whole loop from source of truth to review and frames run through scripts such as `run_comic.py` and `run_spiral.py`. The idea is that a frame does not approve itself. Codex checks the narrative, Gemini and Codex blind-read the visuals, then a deterministic diff finds mismatches in facts, labels and faces. The result is assembled into a single viewer file. A `--dry-run` prints prompts at no credit cost and without an agent.
Who it is for. Researchers and visual storytellers who need a long frame-based narrative without drift in facts, labels and character faces.
Good fit when
- A comic of dozens of frames is needed where numbers and labels must not change from frame to frame.
- A story idea for a presentation becomes a storyboard with locked character looks.
- You study how cross-model review catches mismatches in images.
- The result is needed as one viewer file that opens in a browser.
Not a fit when
- A finished video clip is needed: the project works with images so far and video is planned.
- There is no access to the Codex CLI and a Google reviewer: the loop will not run without them.
- A quick single picture is needed: the pipeline is built for long runs.
Example request
Build a 20-frame comic from this story with locked characters and check each frame with a second model.Limitations
The project has no installer: your agent follows the skills in the repository and the deterministic scripts run from it. Python with the `jsonschema` package, an installed and authenticated Codex CLI, a Google-family reviewer (the legacy `gemini` CLI no longer works, Antigravity with a bundled shim is suggested) and headless Chrome or Chromium are required. Each frame spends credits of image providers. The project works with images and video is only planned. The license is MIT.
How to disable. Delete the repository clone and copied skill folders. Created projects, frames and attempt logs live in your folders and are deleted by hand.
FAQ
How do I prepare the environment?
Install dependencies with `python3 -m pip install -r requirements.txt`, install and authenticate the Codex CLI and a reviewer, then run `python3 cli/preflight.py`.
What does `--dry-run` do?
It prints the frame bake prompts at zero credit cost and without an agent. This verifies the pipeline before the paid part.
Why is a second model needed?
A frame must not be approved by the agent that drew it. Codex and Gemini review independently, then a deterministic comparison follows.
What about the Gemini CLI?
The legacy `gemini` CLI no longer works. Install Antigravity and pass `--gemini-cmd "python3 cli/gemini_agy_shim.py"`.
Can it make video?
Not yet: the project makes a movie from images and video is stated as the next step.
Discussion
No questions or reviews yet.
Related
SkillsРаскадровка в одном скилле: из сценария выходит сетка 3 на 3 с режиссерскими кадрами и таблица таймкодов по сценам.
Skill for Claude Code and Codex: turns two character sheets and a line of story into an anime episode with one fixed cast
The agent scripts, voices and draws a narrated documentary in code, with one continuous camera over photos, maps and cards.
A skill that walks a script to a finished film in stages: it fixes the format, then proposes director, camera and lighting.
Skills and a CLI that take Codex or Claude Code from a novel text to finished manga: script, storyboard, art, lettering and PDF.
A skill that makes your agent film a short product video: concept, edit, sound and render on Remotion or HyperFrames
README in short
The README describes the long-horizon visual generation task, the cross-review method, a quick start, workflows and links to skills. A demo and a Chinese version are included.