AlmondGod/tinyworlds

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1378 stars · 104 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A minimal Python implementation of DeepMind's Genie autoregressive world model, including a video tokenizer, action tokenizer, and dynamics model. It is designed as an educational resource for understanding unsupervised, action-less video world modeling.

Use cases

• understand how DeepMind's Genie world model works
• train an autoregressive world model on game video frames
• learn latent action inference without action labels
• experiment with space-time transformers and FSQ tokenizers
• generate playable video predictions from pretrained checkpoints

When to choose

• you want a small, readable codebase to learn Genie-style world models
• you want to train or fine-tune a world model on your own video datasets
• you need a reference implementation of latent action modeling

When to avoid

• you need a production-grade, scalable video generation system like VEO
• you want a polished application with a UI rather than research code
• you lack GPU resources for training transformer models

Facets

library · maturity active

machine-learning deep-learning llm-training machine-learning deep-learning artificial-intelligence python world-model genie video-generation transformer research educational gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="AlmondGod/tinyworlds")