/ic/ - Artwork/Critique

>>6250496
anon, we're still making breakthroughs in AI research. do you seriously think that our current text-to-image processors won't improve as /g/fags improve their algorithms and techniques for their own models?

the tech is still at its infancy, give it some time. however i will have to say that current ai text-to-image generators are rather finicky, and it cannot yet generate images exactly as you envision it by providing it with details about how the composition should look like or how colours should appear, where the light and shadow should fall onto, temporal consistencies, etc etc. it can only generate images according to your prompts, and even then, the algorithms might even interpret your prompt differently from what you actually asked it to do. not to mention that the makers of the algorithms themselves could just make the algorithms not generate a specific image because it goes against their terms of service and even deliberately sabotage them to prevent copyright issues.