git404hub

what is worlddirector fr?

ppetrichor/worlddirector — explained in plain English

Analysis updated 2026-05-18

65PythonAudience · researcherComplexity · 5/5Setup · hard

tl;dr

WorldDirector generates controllable video simulations of a world, using an AI planning step plus a video model to keep objects consistent even after they leave and re-enter the frame.

vibe map

mindmap
  root((WorldDirector))
    What it does
      Controllable video world model
      Persistent object memory
      3D motion planning
    Tech stack
      Python
      PyTorch
      FlashAttention
      Hugging Face weights
    Use cases
      Research reproduction
      Long video generation
      Camera and object control
    Audience
      Researchers
      Video model engineers

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

what do people make with this?

VIBE 1

Reproduce or extend a research method for controllable, long-horizon video world simulation.

VIBE 2

Generate video where object motion and camera movement are planned in 3D before rendering.

VIBE 3

Study a technique for keeping object appearance consistent when objects leave and re-enter a video.

what's the stack?

PythonPyTorchFlashAttentionHugging Face

how it stacks up fr

ppetrichor/worlddirectoramyxvalen/flash-usdt-senderecho-team-joy-future-academy-jd/echo-infinity
Stars656565
LanguagePythonPythonPython
Setup difficultyhardmoderatehard
Complexity5/55/5
Audienceresearchergeneralresearcher

Figures from each repo's GitHub metadata at analysis time.

how do i run it?

Difficulty · hard time til it works · 1day+

Requires multiple GPUs, specific PyTorch and FlashAttention versions, and large downloaded model weights.

No license information is provided in the README.

in plain english

WorldDirector is a research project that generates controllable video simulations of a world, built to keep track of objects over long stretches of video even when those objects leave the frame and come back later. It comes from a group of researchers and is tied to a published paper, a project website, and a set of downloadable model weights. The approach splits the work into two stages. First, a separate large language model plans out where objects and the camera should move in 3D space over the course of the video, producing a JSON file describing every object's motion and the camera's path. Second, that plan gets converted into 2D controls and fed into a video generation model, which produces the actual footage. A technique called appearance binding helps the system remember what an object looked like, so if it moves off screen and returns, it still looks the same. Using this project requires real technical setup. You need Python, a specific version of the PyTorch library, and a package called FlashAttention, plus the WorldDirector model weights downloaded from Hugging Face. Generating video also needs multiple GPUs, since the example command in the README runs across eight of them. The full pipeline involves several manual steps: planning object motion with an LLM, preparing video segments and prompt files, and then running a generation script with many configuration options. This is not a beginner tool or a plug-and-play app. It is aimed at researchers and engineers working on video generation and world models who want to reproduce or build on the WorldDirector method. The README currently states that the full inference code and the 14-billion-parameter model have been released, with example results and demo videos included to show what the output looks like. The full README is longer than what was shown.

prompts (copy fr)

prompt 1
Explain the WorldDirector inference pipeline step by step, from LLM motion planning to final video generation.
prompt 2
What GPU and library requirements do I need to run WorldDirector-14B locally?
prompt 3
Walk me through how to structure the LLM_plan.json file for a custom video scene in WorldDirector.
prompt 4
Summarize how appearance binding works in WorldDirector to preserve object identity across frames.

Frequently asked questions

what is worlddirector fr?

WorldDirector generates controllable video simulations of a world, using an AI planning step plus a video model to keep objects consistent even after they leave and re-enter the frame.

What language is worlddirector written in?

Mainly Python. The stack also includes Python, PyTorch, FlashAttention.

What license does worlddirector use?

No license information is provided in the README.

How hard is worlddirector to set up?

Setup difficulty is rated hard, with roughly 1day+ to a first successful run.

Who is worlddirector for?

Mainly researcher.

peek the repo → explain another one

This repo across BitVibe Labs

double-check against the repo, no cap.