1. Microsoft’s research division has launched Pix2Gif, an AI model that turns still images into animated gifs in seconds.
2. The model uses a motion-guided warping approach based on text prompts to create the illusion of moving images.
3. Pix2Gif is a research project and is not a publicly available product, but the code can be downloaded and run locally.
Microsoft’s research division has developed a new artificial intelligence model called Pix2Gif that can quickly turn a still image into a gif by taking an “image translation” approach. By using a text prompt alongside the image input, the model spatially transforms the features of the original image to create the appearance of motion. While the model is not a publicly available product from Microsoft, it can be accessed and downloaded for experimentation.
Users can run Pix2Gif locally on their laptop by providing an image and a text prompt to generate a gif. The tool can create a 2-second gif in about one minute, though this process may be faster with a fast GPU. The ability to create gif animations from still images opens up possibilities for future video-like models and potential editing capabilities.
The model was trained using 100,000 animated gifs with appropriate captions, although the source of these gifs is not specified. As a research project, Pix2Gif currently serves as a fun and experimental tool for users to play around with, envisioning how their still images might look with added motion. While it may not be integrated into Microsoft products at this time, the technology behind Pix2Gif could have implications for future AI design and editing tools.