FoundationVision/Infinity
An autoregressive model that generates high-resolution images by predicting discrete visual tokens bit-by-bit using a transformer architecture.
Not currently ranked — collecting fresh signals.
star history
Infinity is a bitwise autoregressive approach to high-resolution image synthesis that scales autoregressive modeling for visual generation. The model uses a transformer-based architecture to predict visual tokens, treating image generation as a sequence prediction problem similar to language modeling. It includes an image tokenizer for discrete representation and supports text-to-image generation, achieving competitive results compared to diffusion-based approaches.
Frequently asked
- What is FoundationVision/Infinity?
- An autoregressive model that generates high-resolution images by predicting discrete visual tokens bit-by-bit using a transformer architecture.
- Is Infinity open source?
- Yes — FoundationVision/Infinity is open source, released under the MIT license.
- What language is Infinity written in?
- FoundationVision/Infinity is primarily written in Python.
- How popular is Infinity?
- FoundationVision/Infinity has 1.6k stars on GitHub.
- Where can I find Infinity?
- FoundationVision/Infinity is on GitHub at https://github.com/FoundationVision/Infinity.