Swatcher

By the way, I should probably go over some of the other features DALL-E 2 can do. I haven't mentioned them yet. Making variations: Upload any image (AI or real) and DALL-E 2 can make variations of it. See this timestamp. Editing / "Inpainting" / "Outpainting": DALL-E 2 can fill in empty transparent areas--Say for example, you can erase part of an image to transparency and describe what you want there, and DALL-E 2 will fill in the erased area. Similarly, you can also crop out/resize an image and DALL-E 2 will fill the rest of the empty transparent canvas. Upload must be a .png file, to support transparency. EDIT: For more on "Outpainting" see my later post. It's definitely one of the coolest, most powerful parts.

Slowly but surely, I'm getting there. 16 male wizard, dungeons and dragons portrait, digital art 17 male wizard, dungeons and dragons D&D RPG, modern digital art portrait ('bolding' = changes or new words) 18 male wizard, dungeons and dragons D&D RPG, modern realistic digital art portrait 19 male wizard, dungeons and dragons D&D RPG portrait, modern realistic digital painting

14/15 "The Statue of Eothas & Statue of Liberty will never be ballin—" EDIT: Part 2

DALL-E 2: THE SECOND BATCH (Only a week left til Sep17. Anyone else, feel free to get in on this. How do we make it look like PoE?) 8 Black woman with white hair, Dungeons & Dragons portrait 9 portrait of a male rogue with curly black hair, in the style of Frank Frazetta 10 painted portrait of a male rogue with wavy black hair and facial hair, in the style of Larry Elmore, Dungeons & Dragons 11 painted portrait of a knight, in the style of Larry Elmore #3 Earth Wind and Fire? Maurice? 12 painted portrait of a knight, in the style of Frank Frazetta I like all these but #3 is wicked 13 portrait of white woman with ginger hair, Acrylic painting, classical fantasy art, Dungeons and Dragons, in the style of Frank Frazetta I just threw like all the words in on the last one ______________________________________________________________________________ I'm realizing this site highly compresses each image to a few kb, so unlike the trash batch I'll be fitting these pics into one post from now on. As expected, it's not Pillars of Eternity, but some of them are pretty cool. Wouldn't be out of place in a completely different RPG. If in case you ever wanted to make art of characters for the classic TTRPGs

Mmm it might pull out some fake video game screenshots, which it has done before (for example I can invent Call of Duty 10: Space Warfare), but it won't ever be Pillars of Eternity I think. It needs to be well-known enough. Another fantasy game, on the other hand... I'm going to try to take some advice and use more old-school fantasy artists, along with more well-known terms like D&D for this next batch.

43 credits remaining. [INSERT COIN] From this trash batch, the biggest issue is DALLE2 doesn't recognize "Pillars of Eternity" unlike DALLE Mini/Craiyon. It's harder to get portraits resembling Pillars of Eternity's art style, without the right keywords. Need to get creative... I'm trying to think of digital/oil painting portrait terms or using other CRPGs. Maybe even other video games. I'm open to any and all ideas for better prompts than mine

Trash batch 7 computer role playing game, player portrait. Are you a WEART gamer? I'm more of a Pleyec Pey person. Laclec Cag is a bit too mainstream for me, but at least their forum isn't filled with racists.

Trash batch 6 fantasy halfling, RPG portrait Well it seems to recognize "halfling". The right one looks like something out of a 2000s flash game on Newgrounds

Trash batch 5 fantasy halfling

Trash batch 4 a halfing in the style of a CRPG portrait Left, I-... Is that Vivec? Is he a coomer???

Trash batch 3 Pillars of Eternity (lmao) Doesn't seem to even recognize Pillars of Eternity as a game. DALL-E Mini/Craiyon does, tho. Repeating what I said, they learned from different training datasets. One AI was fed Pillars of Eternity images, and one wasn't. Big neural networks basically take in a giant feeding-trough worth of data--like in the order of millions--and they're picking what to feed it, which varies. That's why the AIs have different results.

btw, note that the resolutions are higher than they look. Here's a full image from the first one 1024x1024 .png files (converted to a jpeg tho to fit here)

Trash batch 2 portrait of an orlan from Pillars of Eternity Doesn't seem to recognize orlan

In chronological order of how I made them: 1 Black woman with white hair, Pillars of Eternity portrait Tried to see if I could recreate my very first picture in one go. Not bad but the styyyyyle . That sh!t ain't Pillars.

THE FIRST BATCH (7 done, 43 tries left) Alright, this batch was total trash. They LOOK good, but they're the wrong things/wrong style. I need to change the wording/prompts. Take a look: Warning: Long image (had to stitch it together as one long image bc of the inconvenient 1 MB limit. Anyone know a way around that?) EDIT: This website keeps shrinking it down so it's unreadable. I guess I'll just have to screencap & upload it one at a time, post by post, all meticulously. See you guys in a bit

Thanks! Making these is pretty entertaining. Okay we barely even have a week left until my credits reset. Probably should've started earlier. Let's hurry and get some ideas in! I'll start with orlan and halfing, since those were the first two ideas from thelee. This is probably going to take some experimentation. And unlike DALL-E Mini, we can't keep retrying freely (can't use Mini to practice this either, because they're two different AIs with different training data = different results)

I'd be amazed if it did, because DALL-E Mini/Craiyon couldn't at all, although DALL-E 2 was trained on different data sets. Still tho But, I have to do some stuff irl. I'll get back to this soon edit: In the meanwhile, check this out https://dallery.gallery/the-dalle-2-prompt-book/, either for funni pics or to see how a good prompt for DALL-E 2 is made.

UPDATE: I have access to DALLE-2! Man, this thing is cool. However, my credits expire the 17th 18th of every month. They gave me a bunch of starter credits that'll expire by Sep 17 18. I haven't really toyed around with it much, so I still have 50 starter credits (50 'uses'). Every month after, I'll only get 15 free credits (the prev. month's credits don't roll over), and extra costs $$$. If anyone has any ideas, feel free to shoot them now! I can only try so many times by myself. If only there were more DALLE2 people here to try this stuff out and do requests (or one person with a big wallet and a big heart) So, anyone have any ideas for prompts? Portraits?

Bit off-topic but does anyone have access to DALL-E 2 or Midjourney or other good AI models? Just started a thread and I'll repeat some of what I said there: as of now, you can make custom portraits using neural network AI. The AI I always saw used to be hilariously bad at faces, until around this year. Lately, 'DALL-E mini' is the one I've been messing around with the most. Not the best one, but it produces some surprisingly good stuff despite its limitations. AI-generated portraits I posted before, with my edits. Mini's resolution/sharpness is low tho Last one's face is from a real portrait. The rest are 100% from DALL-E mini. Took some keyword kung-fu + edits, but it's all AI, eyyy And one more, in-game. It's also the first one that I made couple days ago: Back to the original question, does anyone have access to a higher quality AI, in particular the betas for DALL-E 2 or Midjourney? Seriously, they're amazing. Google what they can make or check their subreddits in particular. The most impressive models I've seen yet What kind of portraits could someone make with those AI instead?

I tried doing just "Dark Souls" for portraits and got some unexpectedly good results. This was the 1st attempt Half-considering fixing up one of these into a portrait. Man if I had the patience, a lot of these would make a good base for drawing a portrait from scratch. Two more Even though DALL-E mini isn't the best quality this is too much fun. I should probably go set up DiscoDiffusion or anything with Diffusion Models, but I just can't help but love this one's sheer variety and speed. Still though, resolution is meh. If anyone here somehow has access to one of the better AI models, I'd like to see what they can make with it.

Trying anything that isn't "man", "woman", or non-human terms is iffy at best. Currently struggling with "Female Dwarf" and Orlan prompts. Related: After fiddling with keywords, I think I discovered the secret keywords for Coastal Aumaua. The best description of Aumaua:

Even if there's better choices, DALL-E mini is pretty fun to use. Sometimes the AI even decides to have a good day, like on my 1st try this knight dude is just straight up usable An actual face! Minus the snail ear, which is only minor. Just Mike Tyson a better ear from someone. For some reason "knight" alone just produced a lot of decent results. By the way, this guy

A couple observations I made 1. Men seem to be harder to get good faces for. Women usually only take me a few prompts. Maybe it's not as weird as I think. Probably less photos of men than women online, so a weaker sample size to train the AI. Sucks 2. Close-up is good. Especially for faces. The further away & smaller anything gets, the worse the AI's detailing gets (this is also true for other AI models, in my experience. Not a DALL-E thing). Including terms like 'close-up' yielded better results. Try something similar. 3. Faces are the hardest for a lot of AI, so try to use a face from an existing RPG portrait. Of course, as of this year there's better AI models like DALL-E 2 that can do highly accurate faces, IF you have access... Anyway, if you're really gonna use DALL-E mini's faces, focus on getting a good face first, then get the body/armor later. Don't be picky. It's harder to find a face that fits a body than the other way around. Angles are the absolute worst. And speaking of picky, the biggest weakness of this: You can't get too specific about what you want your results to look like, or it'll take forever. I just picked whatever looked good. Don't look for a tanned face at a 45° angle eyes cast down with curly black hair unless you wanna spend five years re-running prompts.

Old pale elf man Seems a little scarred. Might've been the balrog that got his face. Cropped:

Pale elf man in plate armor Armor seems a tiny bit too sharp compared to his softly lit features. Might try flattening it out. And cropped to size: EDIT: Slight armor lighting adjust