The technology race between large companies like Meta and Google is becoming more and more exciting as they continuously introduce new AI applications to meet the growing needs of users in creating images and videos from descriptive text. Recently, Meta launched Imagine with Meta and Make-A-Video, competing with Google AI.
Like OpenAI’s DALL-E , Midjourney , and Stable Diffusion , the new application Imagine with Meta is being developed based on the social network giant’s Emu image generation model.
Notable is the confrontation between Google and AI Gemini and Meta with Imagine with Meta and Make-A-Video . These two apps are both created to turn text descriptions into high-quality images. While Google’s Gemini AI focuses on image generation, Meta offers a multipurpose application, allowing users to create both images and videos with descriptive text.
Imagine with Meta, which runs independently on the web, is built on top of Meta’s existing Emu AI model. Not only does this app generate high-quality images, it also allows creating four different images for each description. This is a big step forward, bringing flexibility and diversity to users.
However, creating tools to create images and videos from text not only brings success but also faces great challenges. One of the main issues Meta is facing is ensuring transparency and avoiding bias in its AI technology, as in the recent case of a racist sticker creator.
To address this issue, Meta has committed to watermarking content created with Imagine with Meta. This watermark is resilient to common edits such as cropping, resizing, and color changes, thereby enhancing image transparency and provenance.
Along with adopting measures such as fingerprinting, technology companies are working to meet regulations and standards related to labeling and authenticating content created with AI technology.
In China, regulations requiring identification marks for AI-generated content have been applied, demonstrating great concern about disputes related to the origin and authenticity of information.
Besides creating images, Meta has also taken a major step forward in creating videos from text with Make-A-Video. This technique, while impressive in its ability to convert text to video, still faces some problems.
Meta’s Make-A-Video creates stop-motion quality videos that give off a strange and surreal feel. Despite progress, the quality of the videos still leaves viewers feeling strange and uncomfortable.
These not only raise questions about the authenticity of the information generated from this technology, but also about the extent to which content is created that is purposeful and aesthetically pleasing. Although an important step forward, improving the quality and fidelity of video from text remains a major challenge for researchers and technologists.
Read more articles:
[visual-link-preview encoded=”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”]
[visual-link-preview encoded=”eyJ0eXBlIjoiZXh0ZXJuYWwiLCJwb3N0IjowLCJwb3N0X2xhYmVsIjoiIiwidXJsIjoiaHR0cHM6Ly80dGVjaHZpZXdzLm5ldC9lbi9ydW53YXktZ2VuLTItcG93ZXJmdWwtYW5kLWZyZWUtdGV4dC10by12aWRlby10b29sLWZvci1ldmVyeW9uZS8iLCJpbWFnZV9pZCI6LTEsImltYWdlX3VybCI6Imh0dHBzOi8vNHRlY2h2aWV3cy5uZXQvbWVkaWEvMjAyMy8wNy9SdW53YXktR2VuLTItQ29uZy1jdS1UZXh0LVZpZGVvLW1hbmgtbWUtdmEtbWllbi1waGktY2hvLW1vaS1uZ3VvaS00VGVjaE5ld3MuZ2lmIiwidGl0bGUiOiJSdW53YXkgR2VuLTI6IFBvd2VyZnVsIGFuZCBGcmVlIFRleHQtdG8tVmlkZW8gVG9vbCBmb3IgRXZlcnlvbmUiLCJzdW1tYXJ5IjoiTm90IGxvbmcgYWdvLCB3ZSB3aXRuZXNzZWQgdGhlIGJyZWFrdGhyb3VnaCBpbm5vdmF0aW9uIG9mIFJ1bndheSBHZW4tMiwgYSB1bmlxdWUgb25saW5lIHRvb2wgdGhhdCB0cmFuc2Zvcm1zIHRleHQgaW50byBzaG9ydCA0LXNlY29uZCB2aWRlb3MgYmFzZWQgb24gdGhlIGluZm9ybWF0aW9uIHlvdSBwcm92aWRlLiIsInRlbXBsYXRlIjoiY29tcGFjdDIifQ==”]