solaris-wm

solaris-wm / solaris

Public

The first multiplayer video world model in Minecraft

solaris-wm.github.io
121
4
100% credibility
Found Feb 26, 2026 at 57 stars 2x -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

JAX implementation of Solaris, a multiplayer Minecraft world model that generates future video frames conditioned on past observations and player actions.

How It Works

1
🌟 Discover Solaris

You stumble upon Solaris, an AI that predicts what happens next when two players adventure together in Minecraft.

2
💻 Set up easily

Create a simple space on your computer for the AI to work, like unpacking a toolbox.

3
📥 Grab ready brains

Download the smart AI brain and fun example gameplay clips from trusted places.

4
▶️ Generate predictions

Click run to let the AI imagine and create videos of future moments in the games.

5
🎥 Watch the magic

Compare real gameplay side-by-side with the AI's predictions – see how well it understands teamwork!

6
📊 Check results

View simple scores showing how realistic and accurate the AI's videos are.

✅ Share your creation

You now have impressive AI-generated Minecraft videos to show friends!

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 57 to 121 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is solaris?

Solaris is a Python/JAX toolkit for training and running the first multiplayer video world model in Minecraft, predicting joint video frames from two players' observations and actions. It simulates shared virtual worlds, like the first multiplayer game ever but for AI-driven video generation. Download pretrained weights and eval datasets from Hugging Face, then run GPU inference to generate videos or TPU training across four stages for custom models.

Why is it gaining traction?

It stands out as the first GitHub project tackling multiplayer video prediction in games, blending world modeling with Minecraft's rich environment—think first multiplayer FPS game but for AI foresight. Developers hook into simple CLI commands for inference on 48GB GPUs, auto-computed FID metrics, and VLM self-consistency evals, skipping boilerplate data prep via sharded HF downloads.

Who should use this?

AI researchers prototyping multi-agent world models for games or robotics, Minecraft AI devs needing predictive rollouts, and RL engineers simulating joint behaviors without full game engines.

Verdict

Promising for cutting-edge research despite 44 stars and 1.0% credibility score—docs are solid with setup scripts, tests cover core paths, but expect TPU-only training and high hardware needs. Fork and experiment if you're in video world modeling.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.