Emu edit is a new Ai based feature launched by Meta which includes Emu Image and Video Generation, Ai researchers from Meta said, today they have made best advancements in AI powered image and video generation. The Instagram and Facebook parent company has developed new tool that enables to edit images via giving text instructions and generate videos by a new method text-to-video generation. These new tools are based on Meta’s Expressive Media Universe or EMU.
Meta CEO said,
Announcing two AI research capabilities built on our foundation model for image generation, Emu. Emu Video generates high quality videos from a text prompt and Emu Edit can precisely alter images based on text inputs. We’re working on integrating these creative editing capabilities into video and photos for Facebook and Instagram.
Meta announced EMU in September and today its being used in production, Meta AI’s imagine feature made powering experiences and this allow users to generate ultra-realistic images in Messenger. In a blog post, Meta’s AI researchers explained how generative ai works step-by-step and generate high quality image from text.
EMU Edit for Image Generation
This AI feature enables user to generate or manipulate their images by giving a prompt, it use EMU edit for doing this task professional. It includes adding/removing backgrounds, colors and shapes/stickers, object detection and segmentation with many more high powered features.
It offers text-based approach to manipulate images and give it a new look that defines your thoughts to others on social media. Meta used this segmentation feature in AI stickers.
EMU edit is designed with 10 million synthesized samples of images dataset. Emu edit delivers high quality and fine output based on the prompt provided by user.
Related Post :
Instagram gives you option to share posts with “Close Friend”
EMU Video for Video Generation
Meta also focused on video enhancing by AI. Meta AI team explained the process of using EMU Video to generate AI videos which is likely similar to Image Generation process. EMU Video tool use benefits of EMU model and provides a text-to-video generation based on diffusion models.
This tool can respond to various inputs, including text only, image only or both together. Video generation process is in two steps, first step is to create an image rendered on the basis of prompt, and before generating a video based on that image add another text prompt.
As team said, this ‘factorized’ approach gives an extreme way to design video generation models. this can be used to animate user’s images based on just a text prompt.
Based on human evaluations, the company’s latest advancements in image generation have demonstrated a significant improvement in both overall quality and adherence to the original text prompt. Human evaluators consistently expressed a strong preference for the newer images compared to the company’s previous work in the field.
Image credits goes to their respective owners