#7575th of 80 in Open modelsFor science, health and research, robotics
dreamdojo
Robot world model trained on 44k hours of human egocentric video
Documentation ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 1.1k
- License
- Apache-2.0
- Last commit
- Mar 2026
- Language
- Python
745 on 22 Apr 20261,132 on 11 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
DreamDojo is NVIDIA's interactive world model for robots, pretrained on 44k hours of human egocentric video and post-trained on robot data. The repo has code for latent action model training, pretraining, post-training, distillation and evaluation. Distilled models are reported to generate autoregressively at 10 FPS for over 1 minute. 2B and 14B checkpoints and GR-1 datasets are on Hugging Face.
Who it is for: Robotics researchers training or evaluating video world models
Strengths
- Covers the full pipeline: latent action model, pretraining, post-training, distillation, evaluation
- Releases 2B and 14B pretrained and post-trained checkpoints
- Includes GR-1 post-training datasets and evaluation sets
- Apache-2.0 source code with paper and project website
Weaknesses
- README gives no hardware requirements; setup details are only in linked docs
- Research code for robotics, not a general-purpose video generation tool
- No tagged releases
- Pretraining on 44k hours of video implies heavy compute, but no figures are given
What it needs
- Docker
- Needs Hugging Face
- Models: DreamDojo 2B, DreamDojo 14B
Also in Open models
See all 80| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 94 out of 100 |
| 2 | layaSingle-pass multilingual classifier returning typed decisions with calibrated odds | 88 out of 100 |
| 3 | rf-detrReal-time transformer models for object detection, segmentation and keypoints | 81 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 78 out of 100 |
| #5 | open_clipPyTorch library for training and running CLIP-style image-text models | 78 out of 100 |