img2img-turbo is an open-source research project that performs image-to-image translation in a single step. Built on Stable Diffusion Turbo and a distilled CycleGAN approach, it can turn a rough sketch into a detailed picture or convert a daytime scene into a night scene almost instantly, where conventional diffusion pipelines would need dozens of steps. The project has earned more than 2,400 stars on GitHub and was last actively updated in August 2025.
Standard image-to-image diffusion is slow because the model must iteratively denoise an image over many steps. img2img-turbo instead trains a generator that produces the translated image in one forward pass, using the distilled SD-Turbo model as its backbone. The repository ships both inference code and training code, along with pretrained models for popular tasks such as sketch-to-image, day-to-night, and night-to-day conversion. Because generation happens in a single step, the approach is interesting anywhere latency matters — interactive drawing tools, live previews, and creative applications where users expect feedback immediately rather than after a long render. The project page links pretrained checkpoints and example results, making it easy to evaluate the output quality before committing to training your own variant on custom paired data.
img2img-turbo is completely free and open source under the MIT license. You only need your own GPU to run or train the models; there is no hosted paid tier attached to the project.
img2img-turbo is for researchers, creative coders, and product builders exploring fast generative image editing. If you are prototyping interactive drawing tools or studying distilled diffusion models, it provides a compact, well-known reference implementation to build on.