The new AI can “reshape” your images in infinite ways.
UK/California based tech startup Stability AI has started Stable Diffusion Reimagine, an image-by-image AI that generates brand new images inspired by an image uploaded by a user – and it will be open source.
The background: A number of impressive text-to-image AIs were released in 2022 – programs that can create images based on text prompts – with one of the most popular examples being Stability AI’s Stable Diffusion.
A major reason for this popularity was that unlike DALL-E 2 and most other text-to-image AIs, Stable Diffusion was open source – users could access the code and create unique models e.g. B. those that only Pokémon or artworks generated their personal style.
what’s new Stability AI has now announced the release of a new tool called Stable Diffusion Reimagine; Instead of generating new images based on text prompts, it creates images inspired by uploaded images.
Stable Diffusion already had a feature called “img2img” that allowed users to upload images along with a text prompt to guide the AI. Reimagne appears to be a simplification of this feature, eliminating the option of a written guide.
“Stable Diffusion Reimagine … allows users to create multiple variations of a single image without limitations,” writes Stability AI. “No need for complex prompts: users can simply upload an image to the algorithm to create as many variations as they like.”
Stability AI has already made Stable Diffusion Reimagine available online and plans to make the code available on its Github page “soon”.
Results may vary: Stability AI lists several use cases for Reimagine, noting that creative agencies could use it to generate options for clients, while web designers could upload a photo to get similar alternatives to use on their websites.
However, based on our initial experiences with the tool, its outputs don’t seem quite ready for such uses – when we uploaded the same source image in the example above, the three images originally generated by Reimagine looked far less realistic and had odd proportions.
The stability AI notes the tool’s limitations and lets users know that they might get some less than impressive results mixed in with the amazing ones, but after half a dozen tries with the same source image we still didn’t get one that looked completely realistic.
The bottom line: Stable Diffusion Reimagine could be a valuable source of inspiration for people who are already somewhat artistic – for example, they could take one of the editions above and recreate it without the wobbly footboard or overstretched curtain rod.
Once the code is released, we may see better performing models trained on narrower datasets – if someone has created such a version only generated bedroom interiors, it might be better to get them right.
In the meantime, there will surely be countless people just wanting to tinker with Reimagine – in which case see what kind Mistake it makes is part of the fun.
We’d love to hear from you! If you have a comment on this article, or if you have a tip for a future Freethink story, please email us at [email protected].