Meta has unveiled a new AI model, “Imagine Yourself,” that can create personalized images from a single photo.
With just one reference image, the AI can generate multiple new images showing a person in different poses, styles, and environments.
The model processes both the image and text instructions simultaneously, making it flexible and efficient.
“Imagine Yourself” can also handle group photos by merging data from multiple reference images, and Meta is exploring future applications like video generation.
Meta presents Imagine yourself
— AK (@_akhaliq) September 23, 2024
Tuning-Free Personalized Image Generation
paper page: https://t.co/gcbkQB5nsG
Diffusion models have demonstrated remarkable efficacy across various image-to-image tasks. In this research, we introduce Imagine yourself, a state-of-the-art model… pic.twitter.com/w92wmIY09E
Though the model shows promise, it’s not without its flaws. Competing models sometimes perform better at preserving a person’s identity, but Meta aims to improve on this.
For now, the technology is still in development, and the model is not yet available to the public. But the potential for creating custom images from just one photo is an exciting step forward in AI.