I’m most excited to use #dalle2 as a tool to bring books to life. I’d love having visual representations of characters and environments to share with others when recommending books. Dalle2 could respark a love for reading for many people #openai
@DrewMedina20 Thanks Drew, I’ve been testing batches for about 4 days now and finally learned how to reduce blur in diffusion renders. It’s been refreshing to take a break from only doing videos. Got a little burned out tbh. Was spending way too much time testing post processing techniques
@NeuroTreats Lol yes, the model’s creator did use it to primary make Princesses, but it’s capable of making beautiful images too. Same ole generative models and CLIP in the background. I just enjoy how simple and fast it is to use in Visions of Chaos. Fun to run batches while at work
@maaartiin_mac Didn’t have to find tune anything. Just kept running prompts for days and learned that the resolution and seed is almost more important than the prompt with this model. About 1 in 20 generated images are impressive. So, it just comes down to mining for the good ones
@marcgregoryart Thank you, and I'm working on one. I got COVID, my wife just gave birth, and I've been testing methods to remove flickering. So, it's been a busy month, but I'll have some time soon to set aside for the Pytti tutorial
@EvaSilvertant Mobile devices (smaller screens) hide a lot of the flickering in videos. If you view on desktop, you'll see differences. As to details lost, that's due to twitter's compression. The method I use smooths out sudden changes which leads to more smooth surfaces -> more compression
I haven't posted a video in a month because I've been focused on removing the flickering in AI generated videos. Here is a preview of the new workflow results. It requires 7 software & 10 post-processing steps. Link to 4k version: https://t.co/N8qBV3nXmu #aiart#pytti#ganart
@DazhiZhong Thanks, and yeah I’m actually impressed by how much flickering twitter’s compression hides. You have to watch the video on YouTube on a large screen to see the differences. Twitter must be using a method similar to the one I’ve found
@TheArtfulEscape Thank you for making an amazing game! I’m working on a video series using key moments (without spoils) as video input into different models. In the videos, I’ll have intro text pop ups recommending the game so as many people as possible get a chance to experience this masterpiece
Playing around with VQGAN tracked videos. Not worth a YouTube upload because the flickering is too strong in 4k. #Pytti Limited Pallet is still the best model for smooth videos, that I can make #AI#GANart#aiart#generativeart
@emmacatnip @brandonXYZW That’s it! The game is so beautiful. Working on a full video series to commemorate highlights (without spoils) as video input through different AI Models