I’ve vacillated between thinking AI art is fascinating and unsettling, but I can’t deny that the Overwatch hero fusion renders I recently came upon on Reddit are really cool.
It is amazing that we can now ask an AI to imagine what Soldier 76 and Mei might look like if they were combined, and it can come up with startlingly creative and cool character animation.
RELATED: Post-Launch DLC and A New Story Trailer Has Surfaced For Nioh 2, Calamity Strikes In The Year 1555
Drew Harrison, a writer, accomplished it when he was one of the few thousand people to have access to Stable Diffusion, an AI picture model that is comparable to other well-known systems like DALL-E and Mid journey.
In the beginning, Harrison gave the AI straightforward directions like “build a hybrid picture of Mercy from Overwatch and Junkrat from Overwatch,” and in some cases, he exerted full influence over artistic elements by adding tags like “4K, extremely detailed, cinematic illumination.”
RELATED: Descenders Announced Winners Of Their 2019 Mod Contest, Featuring Wild Stunts And High Speeds
I wasn’t aware you could be so exact with commands until I played about with the DALL-E Mini online tool.
The brightness of Mercy x Junkrat (Mercrat) and the arm-mounted belly on Moira x Roadhog’s (Moidhog) arm really fascinate and frighten me. Frequently, you can’t count on an AI to produce a fridge-worthy sketch on its first attempt. Harrison informed me that creating AI art is frequently a trial-and-error procedure.
Harrison explained that you should try a prompt, observe the results, make changes, rerun the exercise, and continue until you have the desired result. Harrison prefers Stable Diffusion over other AI art tools at the moment because of its speed.
When rendering time and VRAM are the only creative bottlenecks, he calls Stable Diffusion “lightning quick” and “lightweight,” both of which are significant advantages.
Although it seems like you’ll require a pretty powerful setup to operate it, Stable Diffusion is also the most important AI art tool that is in the process of getting open-source.
However, there are still several glaring drawbacks to AI art. As you can see from Mercrat’s strange octo-hand, AI models continue to appear to be unsure of how to handle fingers and palms.