CODEXFILM STUDIO
EN中文
@LeoGuoLabs GitHub

A COLLECTION OF VISUAL POSSIBILITIES

One story.
So many ways.

A little ink. A cut of paper. A whole new world.
Find a visual language, then make it your own.

An open-source library of video styles and production workflows for OpenAI Codex. Style guides, editable projects, rendering tools and audiovisual QA; capabilities vary by style.

Explore the stylesCreate a film

DRAWN WITH INTENTION. DIRECTED WITH CODEX.

Different materials. A shared imagination.ORIGINAL STYLE STUDY
01 — 18THE STYLE COLLECTION

THE COLLECTION

Find your visual language.

All 18 styles include scene renderers

18 executable production kits, including 1 with a complete editable sample project (Whiteboard Explainer)

Readiness describes reusable tools. New subjects still need authored scenes. Public films, local source bundles and full rebuild verification are separate. Compare source access and rebuild scope →

A style is a starting point. Each includes its own visual rules and tools; sample films and package readiness are listed separately.

Compare capabilities ↗

YOUR TURN IN THE DIRECTOR’S CHAIR

Bring the story.
Make it yours.

Pick a style and tell your Agent what you want to make. It handles setup, production and review; you guide the story.

Read the getting-started guide

Open this repository in your coding Agent and describe your film. / 在 Agent 中打开仓库，描述你想制作的影片。

Use the midcentury-cartoon style to make a film about [subject]. Handle project setup, production and review, then deliver the MP4 and editable project. / 请用 midcentury-cartoon 风格制作关于[主题]的视频，自动完成项目创建、制作与检查，交付 MP4 和可编辑工程。

CODEX FILM STUDIO / FIELD GUIDE

How do you make a video with Codex Film Studio?

Choose a style, author the scenes and review the delivery against the actual scope of the package and sample.

How do I start with Codex?

Open the complete repository in Codex and describe the subject, audience, style, language and duration. The Agent follows the project workflow to create an external project, prepare dependencies, produce and review it. Delivery includes an MP4, editable project and actual review coverage. Start a film →

Can I reproduce a sample exactly?

In principle, yes. How closely the result matches depends on the model’s capabilities and reasoning effort, the completeness of the reference, and the available assets and production environment. Comparing renders with the sample and iterating on them can improve fidelity. Whiteboard sample walkthrough →

What dependencies and inputs do I need?

Use the full checkout, including root scripts, styles and shared methods; copying the Skill directory alone is insufficient. Dependencies vary by style. Whiteboard uses Node, Python and FFmpeg. New subjects need a script, usable assets, recordings or access to a voice service. Dependencies and public inputs →

How do narration and TTS work?

You can import an existing recording or have the Agent generate narration with the included Edge TTS tools, which require internet access but no personal API key. A configured Chatterbox service is another option. Specify the language, voice and speaking rate, then align pictures and captions to the actual audio duration and review pronunciation, pauses and synchronization. For sample replays, reuse the original narration when available. Narration & TTS guide →

Can I install it in another Agent?

In principle, yes. A coding Agent that can read and write project files, run commands and prepare the required dependencies can use the studio. Clone the complete repository and have the Agent read AGENTS.md and the chosen style’s production guide. If it supports Skills, load the repo-local codex-film Skill using that Agent’s conventions. The production and review steps it can perform depend on its tools, permissions and runtime environment. Getting started →

LeoGuoLabs · Documentation updated · Source & revision history ↗