Meta continues to incorporate AI into its services with Emu Edit and Emu Video, two tools for editing images and generating video clips using Facebook and Instagram text descriptions.
Meta, like any good self-respecting tech giant, has also entered the AI race with LLaMA (Large Language Model Meta AI), a language processing model competing with OpenAi’s GPT-4 and Google’s PaLM, but also CM3leon, an image generator competing with DALL-E and Midjourney, or its AudioGen and MusicGen tools. So many technological innovations that the group intends to incorporate into all these services, whether for advertising, to generate stickers in WhatsApp or to introduce a conversational assistant in our favorite instant messaging or on its social networks! And he doesn’t intend to stop there! Also, Meta has just revealed in a blog post two new AI-powered editing tools for Facebook and Instagram: Emu Edit and Emu Video. Based on Emu (Expressive Media Universe), the company’s model for image generation presented last September, they will allow users to edit their photos and produce short videos from simple text descriptions.
Emu Edit and Emu Video: AI comes to Facebook and Instagram
Emu Edit works like any AI-powered image editor: just write your request – “replace the background with city buildings” for example, or “add glasses to the glass” – for the artificial intelligence to run. The demo video resembles tools offered by Adobe (Firefly) and Canva, allowing users to remove or replace objects and people in photographs without having professional image editing experience. However, the particularity of the Meta tool lies in the fact that it only works from a prompt. The user does not need to manually select the part of the image he wants to edit. So, just write “turn the dog into a panda” for Emu Edit to identify the dog in the image. Meta also clarifies that the tool will strive to only make changes that are relevant to the edit request – asking it to add text to a baseball cap won’t change anything else about the design. said cap. With Emu Edit, Meta declares that it wants to offer “a new approach that aims to streamline various image manipulation tasks and bring improved capabilities and precision to image editing”. To do this, the company has developed the Segment Anything Model (SAM), a tool capable of distinctly identifying the different elements making up an image. It’s the same tool that allows you to create personalized stickers on Instagram (see our article).
The second tool, called Emu Video, allows you to generate short videos – we are even closer to GIF than video – from text commands, images, or both at the same time. If the rendering is currently quite unrealistic, the result is however much more convincing than with Make-A-Video, Meta’s AI unveiled in 2022. For this, Meta has ensured that the generation of images is conditioned by a textual request, then that the generation of videos is conditioned by both the text and the generated image.
The company is already dreaming of the possibilities offered by such tools in its social networks. “Imagine being able to create your own animated stickers or smart GIFs on the fly to send to the group chat, rather than having to search for the perfect media for your response. You can also edit your own photos and images without needing to technical skills. Or add some dynamism to your Instagram posts by animating static photos. Or generate something entirely new”, she enthuses. However, no date for the integration of Emu Edit and Emu Video into Instagram and Facebook has yet been revealed.