gnobitab/InstaFlow
InstaFlow achieves one-step text-to-image generation by distilling Stable Diffusion using Rectified Flow, published at ICLR 2024.
Not currently ranked — collecting fresh signals.
star history
InstaFlow is a method for one-step image generation that distills multi-step diffusion models into a single step using Rectified Flow, a continuous-time generative model framework. It builds on Stable Diffusion and achieves comparable image quality to multi-step methods while dramatically reducing inference time. The project provides code, pre-trained models, Hugging Face demos, and Gradio interfaces for local usage.
Frequently asked
- What is gnobitab/InstaFlow?
- InstaFlow achieves one-step text-to-image generation by distilling Stable Diffusion using Rectified Flow, published at ICLR 2024.
- Is InstaFlow open source?
- Yes — gnobitab/InstaFlow is open source, released under the MIT license.
- What language is InstaFlow written in?
- gnobitab/InstaFlow is primarily written in Python.
- How popular is InstaFlow?
- gnobitab/InstaFlow has 1.4k stars on GitHub.
- Where can I find InstaFlow?
- gnobitab/InstaFlow is on GitHub at https://github.com/gnobitab/InstaFlow.