google-research/magvit
A masked generative video transformer that generates videos using a tokenizer and transformer architecture in JAX.
Not currently ranked — collecting fresh signals.
star history
MAGVIT is a masked generative video transformer that generates videos by tokenizing video frames and using transformer-based masked modeling. It achieves state-of-the-art results across video generation and prediction benchmarks including UCF-101, Kinetics-600, and BAIR robot pushing. The official JAX implementation provides training and inference capabilities for the model.
Frequently asked
- What is google-research/magvit?
- A masked generative video transformer that generates videos using a tokenizer and transformer architecture in JAX.
- Is magvit open source?
- Yes — google-research/magvit is open source, released under the Apache-2.0 license.
- What language is magvit written in?
- google-research/magvit is primarily written in Python.
- How popular is magvit?
- google-research/magvit has 1k stars on GitHub.
- Where can I find magvit?
- google-research/magvit is on GitHub at https://github.com/google-research/magvit.