AI image generation is getting crazy good

Back to Bigfoot, Kling did an excellent job of bringing it to life

Summary

A two panel comic. In the first panel, several teens are pointing at an older bald man. The teens are laughing. The old man is sad. They are all wearing ancient robes. The caption for the first panel is “Go up, bald head! Go up bald head!” In the second panel two bears are attacking the teens. The old man is standing watching, very happy.

Gert and Larry from I Hate Fairyland

Original

AI Art Inspired by SDMB Posts - #21 by Spice_Weasel – seeking AI artists to join in the fun.

For shitz and gigglez, I had Vue (Gemini), create a commercial for our business. It only does 8-second renderings, so I had to splice two of them together. It went… OK.

My biggest problems were in making the guy look the same between videos, as well as having the narrator keep the same voice. Not too sure I was 100% on either of them, but whatever. Close enough.

Was that based on anything, or just a random thought ?!

It’s from the Bible.

Ah. Thanks.

I recognized it! Was it discussed here on the Dope before? I can’t remember where I learned about it.

It was mentioned in the Biblical trash talking thread just a week ago.

Thanks!

Interesting article on Google Gemini’s latest update

https://www.techspot.com/news/109259-google-gemini-nano-banana-ai-image-generator-blowing.html

I figured I’d give Gemini a try after all the breathy reports of throwing a half dozen jpgs at it from across the room and it making amazing images. I started out giving it reference images for three Pathfinder characters from out current game and asking for them to be posed in a shrine with Spruce kneeling and watching a butterfly land on his weapon, Thrakka cooking something over a brazier and Zophiel reading a book.

The result wasn’t bad! I only had one real complaint and it was Spruce’s starknife had two of its three blades in a straight line. I wanted them to be equally spaces and radiating from the center. Gemini sagely agreed and rendered Spruce fighting a metal octopus

I thought maybe some more assistance was needed so drew up a quick outline of what I meant and offered a guide as to where I was talking about. Gemini agreed that it totally understood me now and put a mangled starknife in the upper corner of the image.

At this point I started over. I changed Thrakka to be sharpening his sword. After a very cartoony first attempt and my request for a more classic fantasy art style, it gave me this where it really just copy and pasted Spruce from the reference without any attempt to modify him or his pose.

I asked for his head to be turned and Gemini obliged. The image came out… okayish.

I mean, definitely impressive for a computer to use those references and make this[1] – don’t get me wrong – but certainly not living up to expectations based on the breathless hype to spend an hour arguing with it. Not shown are the five instances where it said “Oh yeah, I get it. You want me to [restates me requests]…” then just gave the identical image to the previous one and said “See, I fixed it!”

Anyway, super cool but less carefree than the articles suggest (not just the linked one above, I’ve seen a lot about this the last two days)


  1. 2022 me would be amazed as all hell to have gotten this, how quickly we get jaded ↩︎

Testing Gemini with a little Mayberry humor. Included my cat for good measure.

Barney was reluctant to try Aunt Bee’s homemade pickles, but she insisted.

The cat is certainly alarmed.

Something about the people in that Mayberry scene makes them stand out unrealistically from the background. It’s like they’re greenscreened in, or something.

From a distance, I only see it on Barney, and I think it may be the color. Closer up, I see the characteristic “opaque edges” that aren’t blended in, which you see on low resolution photoshop jobs.

That’s because of the starting images I fed Gemini. The original had Aunt Bee and Andy. I switched out Andy for Barney, but his image was higher resolution. Gemini didn’t account for that. Could have fixed it, but just kept the first iteration.

Sora did a bit better. Made some mistakes (like Aunt Bee’s right hand), but did a better job of harmonizing/blending.