Search results for query: *

https://labs.openai.com/s/RXTFgWX5DmjvAU9CgHM7uXgX Prompt: Touma Kamijou from A Certain Magical Index punching Cosmic Fear Garou from One Punch Man in the face with his right hand. Yusuke Murata and Kiyotaka Haimura style Did not work unfortunately.

https://labs.openai.com/s/4hpIU6wFXisnZNopIQeE5MCw Prompt: A small girl with long silver hair and green eyes in a white nun's outfit lined with gold with a long hood praying Parmigianino style Okay now this is scary, it's almost to good.

https://labs.openai.com/s/BsOramAQcwIFicrDAbjkJfX5 https://labs.openai.com/s/k7uTtIdlzU1yCNEC7uPXiuEM https://labs.openai.com/s/LMclX2olNs8NC5YUq5ftcJ6P https://labs.openai.com/s/MyP01m2uJ9C08ZdXsBuBI4Xy Prompt: Colored manga art of a 15-year-old boy with long and curly dark brown hair walking down a busy street with a bag of groceries in his right-hand, Michelangelo style.

So I got my hands on Dalle-2 here's some stuff I generated with it. https://labs.openai.com/s/2ibEx0VvPizQ44VbvS2xhdnZ Prompt: Jesus Christ as a 15-year-old manga character in Michelangelo style, black and white A pretty good pic I got from DALLE

For these images I played around with init-images. Original Image: Damn! That looks cool as fuck! Like a more metal version of Godzilla Earth. Also tried to make a black and white version, but it didn't work. It didn't work. But looks cool anyway. Also tried it with Sage Centipede. Original Image: Looks as if Sage Centipede just busted through a wall, there's also a bit of color here and there but the AI wasn't able to finish coloring it before the generation was done. So I did it again with a few changes to the prompt. And setting such as changing the number of steps skipped. I bit more color, but not a lot. If I really extended the number of steps high, like 1000 steps it could work. However, I would need to adjust the skip_steps as well, and if I put the skip_steps too low the picture wouldn't really follow the init_image as that would be too blurry. But too high and a problem like this would occur. Also, the amount of ram and time it would take would be...

Only one image today.

Thought I should do this if other people want to do this kind of stuff. So this is how to use Disco Diffusion 5. First you wanna search it up on Google or click this link: https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&cad=rja&uact=8&ved=2ahUKEwiZu6i3n6_2AhUsU98KHayQCy8QFnoECAYQAQ&url=https://colab.research.google.com/github/alembics/disco-diffusion/blob/main/Disco_Diffusion.ipynb&usg=AOvVaw1DGrjjRabdLpbwFl5WnPTF Once you see the link, click it and it will open Google Colab. You should see this. Now you can either make a copy, or use it like this. I made a copy since using the Colab like this doesn't actually save your previous prompts or settings. If you want to make a copy go to file and click save a copy in drive. Once you do you should have this. As you can see on the top it saves the Colab will not save anything. You can change this by clicking Notebook Settings in yellow. Which will open this, uncheck the Omit code part. And then press save...

So I got Disco Diffusion 5, had a bit of trouble setting it up but I managed to make it work. Here's something I made for a friend. God that looks so cool, DF5 is already better than I thought. Then I tried to use initimages. Original Image: DF5: ... WHY IS HE SQUISHED? HE LOOKS LIKE MINETA NOW! Plus that face is nightmare fuel. I tried again with another image but. Original Image: DF5: Still squished, and his face looks disfigured. It might be because of how big the images are, I'll try using smaller images next time. Also, I did repeat over on Nightcafe to see if it could do it better. Hm, now that I look at it, DF5's and Nightcafe's Red Moon art is super similar, could it be because it's the same prompt? Oh god he's horrifying. And finally, something my friend generated.

Okay so I decided to see how good GLIDE is at human faces. The results. Not very good. Uh this isn’t Leonardo DiCaprio. This seems to be Venice. This isn’t Leonardo DiCaprio either. After this I thought maybe it would do animal’s better? GLIDE seems to actually understand animals. But even then it isn’t perfect as for some reason the bunny’s head is fucked up and their’s red marks. The AI put those red marks there for some reason. I did the same with CLIP-GLIDE it also failed spectacularly. So it seems that while the GLIDE AI understands animals and objects, it does not understand humans and the names of humans. It must of not been trained on that sort of dataset. It also doesn’t seem to understand what Manga illustration meant. The final CLIP-GLIDE art I did. One thing I like is that the dragon doesn’t look like a typical dragon but more alien. Though it’s gray like an untextured model. Probably my fault for making the prompt low poly render. And...

So here's some more arts I generated. Using the Google Colab version of Disco Diffusion. This is my favorite piece of AI generated artwork yet. It seemlessly combined the two art styles together and made this cool ass piece, like a Titan lowering his head to the earth, so tall that the only thing you can see is its nose. I really want this to be in a game or something so I can look around! Though it wasn't what I actually wanted to generate. Really cool though! Made it my wallpaper. Also I'll give you some reputation if you guess what the prompt is referencing. I also had used the same prompt on GLIDE. ...Nice flowers but not what I wanted or expected. I don't think it knew what the artist names and tag meant. I also did it on CLIP GUIDE. ...Better, but still not what I wanted. It seems that while it may understand the actual prompt the GLIDE AI doesn't understand the artist's names and tags mean. Also I disliked how small the images GLIDE generates are, so I used...

Alright so I got my hands on another AI art gen software called GLIDE made by open AI it's a better version of DALLE. Here's a video on it. And here's something I've generated with it. GLIDE seems to be better than Disco Diffusion and faster too.

So thanks to Baughn from SV I got my hands on Disco Diffusion. And decided to make an art thread just like them. See Baughn's AI art thread here: https://forums.sufficientvelocity.com/threads/baughns-ai-art-thread.92461/?post=20770699#post-20770699 Here's what I've generated so far. I also...