The open source community is doing more with GenAI and they're doing it better. If Gimp weren't in such bad shape, Photoshop would be over. This exact feature would already be ready to go and would be far superior to Adobe's version.
Any startup with minimal effort and capital will be able to duplicate this product. They'll eat Adobe's thick margins away.
That said, OSS is in way better shape in the image space than in the LLM space. There are actually products and installers based on SD out there, something that won't happen in LLM space for awhile.
Only been saying this for 15 years or so.
And yet, here we are.
Even Godot is going with that playbook like Blender and doing what GIMP never did. It’s a shame imo.
Imagine a world if GIMP and other open source creative tools had acted this way instead of the status quo of "Patches welcome" and "NOFIX", the attitude in the 21 year old "Adjustment Layer" feature request comes to mind...
but thinking you can fight adobe's photography setup (lightroom, ACR, photoshop, bridge etc) with minimal effort and capital to gain users for 6 bucks a month is delusional. but hey, if it happens, awesome.
Stable Diffusion is extremely cool and is a great project, but it does not compare favourably to Midjourney[1] or Adobe's offering. On the LLM side, StableLM is terrible compared to competitors. Some other "open source" models aren't open source at all, but instead were the "this doesn't compete so let's just dump it in the open" reactions.
Add that Adobe has a moat of legally licensed content that they used to train their model, which might turn into a massive differentiator if penalties fall hard on competitors.
[1] - A lot of people speculated that Midjourney used Stable Diffusion, but the products have veered so completely away from each other, the former just dramatically better for almost any prompt, that if they started with it they have made enormous improvements.