Microsoft Research Asia, in collaboration with Peking University and Microsoft Azure AI, has recently developed a new AI visual generation technology, “NUWA-Infinity: Autoregressive over Autoregressive Generation for Infinite Visual Synthesis”, which has generated a lot of buzz, enabling applications such as automatic image generation from text or learning images to generate expanded images.
Previously, OpenAI’s DALL-E2, Google’s Imagen, and Parti could generate images directly from the text, but the resolution was only 1024×1024, while “NUWA-Infinity: Autoregressive Generation for Infinite Visual Synthesis” could generate high-definition images up to 4096×1024. Over Autoregressive Generation for Infinite Visual Synthesis” can generate high-definition images up to 4096×1024 level.
In addition, the technology can learn from existing images to expand them into larger and clearer images following the same architecture, as well as video conversions, among other things, with five specific capabilities.
Make images move (Image Animation)
Generating video from the text (Text-to-Video)
Text-to-Image
Image Outpainting
Image Generation