|
Doctor Zero posted:I hesitate to put this in the Solo RPG thread, but I was playing with Claude and it's much better at admitting when it doesn't know something, although not perfect. When it does make up something, it's pretty true to the source material. If you've been disappointed with Chat-GPT for gaming, maybe try it. Thanks for sharing these, I haven't used Claude at all and these results are pretty intriguing!
|
# ? Jul 20, 2023 19:35 |
|
|
# ? May 2, 2024 05:15 |
|
Every indication is that SXDL is going to be a massive upgrade over SD1.5. The only "negative" I've heard is that it doesn't have any porno in it, which is causing all the usual freaks to scream about how it's going to be failure. I actually love that, because a couple of times I've asked for, like, a middle aged female monk and got a great gen except they were topless and anatomically correct. At that point it's either crop it or toss it and that sucks. edit: also regarding SD1.5, if the "battlemap" LORA is the one I'm thinking of the guy who makes those absolutely sucks at making LORAs. No real surprise there. He has like 20+ LORAs and they all suck. Megaman's Jockstrap fucked around with this message at 21:32 on Jul 20, 2023 |
# ? Jul 20, 2023 21:27 |
|
I recall reading somewhere that SDXL was also going to substantially increase the system requirements for generating images, to something in the 30XX range. Hopefully that doesn’t prove to be the case, but it wouldn’t surprise me.
|
# ? Jul 20, 2023 23:35 |
|
Kestral posted:I recall reading somewhere that SDXL was also going to substantially increase the system requirements for generating images, to something in the 30XX range. Hopefully that doesn’t prove to be the case, but it wouldn’t surprise me. It is higher but it's not that high. quote:needing only a Windows 10 or 11, or Linux operating system, with 16GB RAM, an Nvidia GeForce RTX 20 graphics card (equivalent or higher standard) equipped with a minimum of 8GB of VRAM. Linux users are also able to use a compatible AMD card with 16GB VRAM. https://archive.is/achPx#selection-1481.113-1481.370
|
# ? Jul 20, 2023 23:46 |
|
I gave another pass at the OP 2.0, the guides and some of the other stuff people posted I think should probably just be reposted whenever we make a new thread.quote:AI tools can be a great resource for generating new content for traditional games like Dungeons and Dragons and Magic the Gathering and other TTPRGs. These tools can help you come up with new spells, feats, magic cards, and more, quickly and easily. Below is a list of tools that might be useful, be aware that there is a debate regarding the ethicacy of their training data, and the broader societal economic effects. To learn more you can go to the AI Debate thread in D&D. And to learn more about the technology behind AI like Machine Learning you can go to the Caverns of Cobol thread here. If anyone has written any neat write ups or more specific guides, if you can make a link to the post I can add it to the Guides section. I'll wait a couple of days and then DM Leper for guidance on whether they should post it, or if I could, or if someone else wants to etc.
|
# ? Jul 21, 2023 01:32 |
|
Megaman's Jockstrap posted:edit: also regarding SD1.5, if the "battlemap" LORA is the one I'm thinking of the guy who makes those absolutely sucks at making LORAs. No real surprise there. He has like 20+ LORAs and they all suck. Probably, since I only know of two and I think this one was a merger of those two, this one this one. I always did suspect I could do a better job at it. The obvious way to improve it would have been to train it on battlemaps with a grid where the grid size and relative size of everything in the map was all the same. But I never did because the major flaw with it all seemed like, once you get it trained enough it cant deviate from the style of the map. You train it on maps in a forest, you get a model that can make maps in a forest. You can train it on a bunch of different things with different tokens and probably get a bunch of different styles but that's a lot of work and once I saw that midjourney could just do it easily I kinda lost my motivation for it. I never had much luck with just vanilla 1.5 or 2.2 either. But now it looks like SDXL might be as good as MJ at it (though I cant say for sure since I've only done a few tests and I never really used MJ for much of anything, just saw other people doing it. It definitely is good at it though.) My next plan is, what if I give img2img a grid, will it keep everything perfectly spaced out on the grid? What if I draw a quick little map over the grid, just the lines where I want the walls to be, will it work? I suspect it will. I think this is a bigger deal for traditional games than being able to do character portraits and stuff. Character portraits are cool flavor but we all need battlemaps and they're a lot of work if you wanna make your own for something you use up fast unless you want to just use graph paper or, what I think everyone does, you just go rip some off the Internet. The one thing I wanna figure out but I cant think of an easy way to do is how to generate modular maps and have the whole process be AI from start to finish? You could give it maybe a skeleton of modular maps and run it through img2img but then, how do you make the skeletons with AI? I got a friend learning to work with tensorflow right now and from what he's been saying it's actually not that hard so a "make outlines for modular maps with AI" project might be possible. I just got too many other projects right now to join him. Megaman's Jockstrap posted:It is higher but it's not that high. It is definitely slower though, substantially slower. I'm using comfyui for SDXL stuff so it's not an apples to apples comparison here but I get about 20 it/s on a 4090 for 1.5 in auto1111 and, a brutal 1.18 s/it (not even it/s) for SDXL in comfyUI. I've heard the dev version of auto1111 is faster with SDXL than comfyui but not 20+ times faster. I think it will be faster once 1.0 comes out because I don't think any optimizations have been done on 0.9 but I doubt it will be as fast as 1.5. I can already tell SDXL is gonna be a ton better for deforum projects, its ability to stick to the prompt and give consistent results is gonna make it just way better for animation. But if it doesn't get much faster then it's gonna be pretty rough sitting there waiting 5 or 10 minutes just to be able to see if that run is going well. Though maybe that will encourage people to work on some of the optimizations that exist in theory but no ones implemented yet and make even 1.5 and 2.1 faster for everyone, I dunno BrainDance fucked around with this message at 02:44 on Jul 21, 2023 |
# ? Jul 21, 2023 02:14 |
|
Raenir Salazar posted:If anyone has written any neat write ups or more specific guides, if you can make a link to the post I can add it to the Guides section. I'll wait a couple of days and then DM Leper for guidance on whether they should post it, or if I could, or if someone else wants to etc. Looks great but I noticed the VN Style Cutscenes! section is missing its images.
|
# ? Jul 21, 2023 03:18 |
|
BrainDance posted:It is definitely slower though, substantially slower. Yeeeeep, this is what I was afraid of. Looks like those of us on pre-30XX cards are going to be sticking with 1.5 for a while.
|
# ? Jul 21, 2023 04:41 |
|
Nvidia has completely bamboozled me with how they change up their series nomenclature, so I had to go look up on Wikipedia. It seems I am two generations behind the minimum reqs, heh.
|
# ? Jul 21, 2023 09:40 |
|
I've been generating things we see in our game on the fly with SDXL now, even though it's slower it's easier because, since it sticks to the prompt I don't have to tweak it for as long. These are still just examples of quick SDXL outputs without much time spent tweaking things to get it good. It's the new DM's game who's just sorta learning, but it got a lot better once we finished the campaign he was following in the beginning and started doing something he's making himself. This is what I got from last night's game First, I tried to make a sigil for Barovia. A blue orc guy we ran into, so this is interesting because SDXL, 0.9 anyway, still messes up hands and feet. I thought that may have been because it's an orc and it gets kinda confused compared to normal humans but then, there's a thing where there's this disease causing people to rapidly age, and we had to deal with a dwarf with it and, yeah, hands are still kinda messed up These are, again, real quick attempts so there might be ways to get it to not mess them up. Maybe those long negative prompts would actually work in SDXL and wouldn't just be a big placebo like in SD 1.5? My character, a clown fairy, currently has this wooden puppet curse. The 1st one needs to be cleaned up but could be a really good image with just a little bit more work. SDXL has two different models, the base and the refiner. It generates a base image with the first one then sends it to img2img with the refiner model to make it "better" Here is the base image for this one and this is it after the refiner. These ones weren't right, I'm a clown but not this kinda clown, but they were still cool I'm under no illusion that these are good, for one I barely know how to use comfyui and I kind of hate it, so I'm pretty much limited to normal txt2img and then the refiner. But, they're much better than I would have gotten for the same effort or probably even a lot more in SD 1.5 The refiner does less than I expected and in one case ruined a good image, though for most it still does make them a little better, but I think that must increase the total generation time by quite a bit since it's basically running img2img on every image. I think what might be interesting, once people understand how it actually works, finetuning the refiner model. It just really seems like a refiner step to check for and fix hands could totally be a thing and I dont get why they're not doing that already with it. Maybe they are and this is as good as it gets? But still, might be a lot of cool things that can be done with it. Kestral posted:Yeeeeep, this is what I was afraid of. Looks like those of us on pre-30XX cards are going to be sticking with 1.5 for a while. I don't think it will be that bad when 1.0 comes out. People keep saying the model currently isn't distilled, and so that might be a huge performance hit? I'm not sure. Also I havent tried to optimize anything. Current nvidia drivers are actually really negatively affecting AI performance. Nvidia did a thing where they aggressively move things in vram to system ram. This is probably good for game performance, really really bad for AI performance. 531.79 was apparently the last one not affected, I havent downgraded since I'm very lazy but that might account for a little bit of it, too. My automatic1111 setup has all the normal 4090 optimizations. Without those I get something like 5 it/s, so that can really account for a lot. BrainDance fucked around with this message at 02:29 on Jul 22, 2023 |
# ? Jul 22, 2023 02:00 |
|
The Eyes Have It posted:Looks great but I noticed the VN Style Cutscenes! section is missing its images. Weird it seems to work fine on my end, anyone else can confirm?
|
# ? Jul 22, 2023 02:34 |
|
Raenir Salazar posted:Weird it seems to work fine on my end, anyone else can confirm? They dont load for me either
|
# ? Jul 22, 2023 02:59 |
|
SDXL being released today. The maker of the RPG model has already said that he's going to bring it over and he has a year of experience with finetuning (that's basically an eternity in this biz right now), so expect the "AI generated RPG art" space to just absolutely explode in quality in about three months.
|
# ? Jul 26, 2023 19:12 |
|
Trip report: out of the box SDXL is far more coherent and far more capable of hitting particular styles and looks without searching for "the magic prompt". This was about 4 minutes of work, trying to make Dark Sun art. You simply wouldn't get this kind of thing out of 1.5. This will be an absolute monster when it has it's ControlNets and LORAs sorted out, which will fix the weird hands and anatomy stuff.
|
# ? Jul 29, 2023 07:00 |
|
poo poo, that hits the Dark Sun aesthetic beautifully.
|
# ? Jul 29, 2023 21:10 |
|
That is straight-up perfect Dark Sun art. Could you post the settings and prompt for that? As soon as the D&D LoRAs / models are out for SDXL, I’m gonna use it to make so many goddamn thri-kreen.
|
# ? Jul 29, 2023 22:15 |
|
Sure. This is on the base SDXL on SD_NEXT/Vladomatic. I actually had a misconfiguration with my VAE that made the image much lower contrast than it should be, but it just happened to work out and make it more Brom-y. You'll notice that there is an artifact extending out from the figure's right hand, but it was very easily masked and painted out. Also: you'll notice that you don't need ridiculous cargo cult long-rear end prompts. This was about my sixteenth try (I have a 4090 and am decently experienced so it goes fast). Also also: the model understands the phrases: "the art of athas and dark sun" "the dark sun of athas" "amazing d & d dark sun art" ...but I didn't use them, as it had a (to me) negative effect on coherence. Ok, enough yappin'. Here you go: Positive prompt: fully body oil painting of A wanderer with pale white skin and warpaint, leather girdle, no pants, sandals, alone on the wasteland, orange sky, (by Gerald Brom: 1.1) Negative prompt: vegetation, plants, water, river, rain, clouds, nsfw Steps: 75, Seed: 2484100468, Sampler: UniPC, CFG scale: 7, Size: 1304x784, Parser: Full parser, Model: sd_xl_base_1.0, Model hash: be9edd61, Refiner: sd_xl_refiner_1.0, Version: 42c6147, Pipeline: Diffusers, Operations: txt2img, Token merging ratio: 0.5 I have a few other things to talk about here in my next post. What it basically boils down to is that SDXL is a large step forward and substantially closes the gap that had been growing between Midjourney and Stable Diffusion. In six months, when they've got ControlNet integrated and we have some additional finetunes this is going to be phenomenal. Megaman's Jockstrap fucked around with this message at 00:17 on Jul 30, 2023 |
# ? Jul 29, 2023 23:02 |
|
Thanks for the rundown, Mega! Speaking of which, turns out ComfyUI handles SDXL enormously better than A1111, to the point where I can run an image in under a minute instead of Considerably More Than That, so I gave it a shot. The interface is clumsy, making changes to your settings is slow and awkward compared to A1111, and my results are nowhere near as good as Megaman's - that VAE misconfiguration was some secret sauce - but it does run, so there's hope for those of us on lower-end video cards. The downside is that ComfyUI appears to load the entire goddamn base model and refiner into RAM to accomplish this feat, so if you don't have at least 16GB, you're out of luck with that method. For folks wondering about battlemaps: Yes. SDXL does battlemaps, although not quite as well as I'd wish: These are not pretty battlemaps, but they are literally the first two I generated, with an extremely basic prompt of "a battlemap of a field of crystals and small pools" for the first one, and, for the second: "a battlemap of a field of crystals and small pools, in the style of (Gerald Brom:1.1), on old parchment paper." If I were running a session on Roll20 and I needed a map right now for an encounter I hadn't planned on, I'd be quite happy with the second one. Now, when I tried to make them actually look good, that's when it started taking a while, and in fact I still haven't managed one that actually does what I want it to. Adding aesthetic keys really throws it for a loop. It looooves generating cartographic maps and isometric maps though, and when you insist that it not do those things it gets rather cross. Which is really Stable Diffusion in a nutshell: it's very, very easy to get either mediocre results that are in the ballpark of what you're looking for, or great results when you have nothing in mind but a vague concept and you let the machine dream wildly, and it's quite time-consuming to execute on a specific vision. There's a lot of potential here though, looking forward to Megaman's next writeup. Edit: Not sure I agree with the "you don't need long prompts" idea. Long negative prompts, yeah, not as much, but if you've got specific concepts to execute, it definitely still needs you to know how to render instructions to it in a way it understands. Kestral fucked around with this message at 05:51 on Jul 30, 2023 |
# ? Jul 30, 2023 05:19 |
|
I was getting much better battle maps with SDXL but I'm not sure which part of my prompt was doing the work. My prompts though specified heavily that it was DnD. I dont remember off the top of my head but it was more "A DnD Battlemap of a fantasy dungeon, top down perspective, birds eye view, Dungeons and Dragons", stuff like that. These are what I'm getting, I dont have the exact prompt for most of them but it's like "A (DnD Battlemap:1.25) in a **, top down perspective, grid battlemap, Dungeons and Dragons, 4k, masterpiece" and I try with some other things. Some of them get a little wild and I think that comes from me trying to put the classic "intricate" in there which SDXL for a battlemap interprets as "walls everywhere" The, I guess, logic is broken for some of them like completely closed off rooms. That I believe could be fixed easily in img2img. Though, I'm doing this all in comfyui and I havent really learned how to use it yet. Auto1111 is garbage right now for anything SDXL, and I'm having problems getting SD.Next to see my GPU. I'm using dpmpp_2m for my sampler. Even though it's slower I haven't been able to get nearly as good results with euler. They need some more work to be usable but I don't think it's much work. And I feel like when I get some better prompt stuff worked out I might be able to get more usable ones. Every now and then though it pops out something really good which is why I really should be writing my prompts down somewhere. I dont think comfyui puts it in the metadata like auto1111 can. BrainDance fucked around with this message at 06:46 on Jul 30, 2023 |
# ? Jul 30, 2023 06:27 |
|
Get some actual battlemap chunks from a module and send them to the interrogator, find out what SD thinks the prompt was. You can get all sorts of fun and cool stuff out of this, and I never see it mentioned as the #1 way to prompt engineer once you get a basic handle on how to construct a prompt. There's no need to do the 4k 8k masterpiece stuff with SDXL, btw.
|
# ? Jul 30, 2023 07:34 |
|
Is there a way for those battlemaps to specify like a DPI so the grid from Foundry can overlap?
|
# ? Jul 30, 2023 13:52 |
|
Raenir Salazar posted:Is there a way for those battlemaps to specify like a DPI so the grid from Foundry can overlap? Throw them into irFanview and adjust the DPI there
|
# ? Jul 30, 2023 16:34 |
|
Humbug Scoolbus posted:Throw them into irFanview and adjust the DPI there Is this a thing!? Maybe it will let me fix any random battlemap I find without a displayed dpi?
|
# ? Jul 30, 2023 16:59 |
|
Raenir Salazar posted:Is this a thing!? Maybe it will let me fix any random battlemap I find without a displayed dpi? Absolutely it's a thing; check the bottom left of the dialogue box. The command is in the 'Image' menu edit; Incidentally, Roll20 uses 70 DPI for it's default. Humbug Scoolbus fucked around with this message at 18:39 on Jul 30, 2023 |
# ? Jul 30, 2023 18:31 |
|
Anyone here figured out how SDXL has changed the way prompt weighting works? The old (term:1.5) or (term:0.7) etc. technique is... Finicky, now, and I strongly suspect I'm missing something here.
|
# ? Jul 31, 2023 04:33 |
|
Use SDXL to make unsettling crime scene photos for your favorite 1930s Lovecraftian game. Makes a great landing page.
|
# ? Jul 31, 2023 07:56 |
|
Megaman's Jockstrap posted:Use SDXL to make unsettling crime scene photos for your favorite 1930s Lovecraftian game. Makes a great landing page. Oh hey that's pretty neat. Also I just wanted to share this
|
# ? Aug 1, 2023 00:07 |
|
The Eyes Have It posted:Oh hey that's pretty neat. Oh my god I feel this so hard. My first real SDXL project involves a character “wielding a sword” and out of the hundreds of images generated, not a single one was anywhere close to correct. Swords emerge from your forearm, right? Or float next to you, but in a derpy way and not in a cool Alucard way, yeah, that’s the ticket! Someone should shove a bunch of German and Italian fighting manuals into a LoRA and see what happens.
|
# ? Aug 1, 2023 00:25 |
|
Has anyone found especially good checkpoints for landscapes in an illustrated/painted style? The "discussion" in the Industry thread gave me an idea for how to solve an especially tricky conversion of an ink sketch into a photo/painting using Regional Prompting, but I'm finding that my checkpoint collection has a real weak spot in the landscapes department. I did attempt a photorealistic style with Photon, which looks gorgeous when it works, but the ControlNet references make it choke up badly. Still, highly recommended for photorealistic landscapes when you're not trying to do something silly like this. Also stumbled across this: https://www.youtube.com/watch?v=ckyhwy0Aui4 Lovely comparison of a ton of artists' styles as recognized by SDXL. For someone like me who has an idea of what they want something to look like, but often struggles with describing it in a way the model will understand, this is a phenomenal reference.
|
# ? Aug 6, 2023 04:07 |
|
Kestral posted:Also stumbled across this: Here's a big visual one (not a video) sort of along similar lines: https://rikkar69.github.io/SDXL-artist-study/ Got that tip from the AI Art Weekly newsletter (which is a wonderful resource in itself)
|
# ? Aug 7, 2023 00:14 |
|
There's some pretty rudimentary AI video generation out there, and so in my quest to make my games as cinematic and luxurious as possible I decided to dive in. I discovered that there are 2 major platforms: Gen-2, that you have to pay for (yuckarooni) and Pika Labs, which is free. Also Pika Labs works with images you upload, while Gen-2 just uses them as a guide and produces it's own thing. I decided to see how it works. The answer is, pretty decent! Here's a Frank-Frazetta style orc warrior that I generated. Please note that the actual output of Pika Labs is an MP4 (I had to convert this to GIF to host it). Sadly, Foundry can't accept these as valid character sheet pictures. I desperately wanted Hogwarts style character pics for my players, even though they don't loop cleanly. Click for big, and keep in mind this is just an SDXL image that I sent over to the algorithm with the description "orc warrior in a rainstorm, moving clouds, rain, lightning, breeze, storm": This tech is like 8 months old so I'm sure it's going to improve rapidly. Pretty cool.
|
# ? Aug 9, 2023 05:25 |
|
Goddam, that's a cool tool! Here's one I made. A little clunky, but incredible to think I provided a basic prompt (and a source image, of course) and blammo.
|
# ? Aug 9, 2023 08:12 |
|
Yeah, it's a cool toy. If they can get it up to 10 seconds or so and an option for seamlessly loopable (a tough ask) they could make a lot of money; that's something I (and I think a whole bunch of people) would actually pay for.
|
# ? Aug 9, 2023 16:08 |
|
Does Pika let you use a prompt and also maintain the original composition? Runway lets you do either/or but not both. But yeah, in a couple of generations this is going to be a Big Deal if it can maintain coherence and be "directed" to any significant degree.
|
# ? Aug 9, 2023 16:12 |
|
feedmyleg posted:Does Pika let you use a prompt and also maintain the original composition? Runway lets you do either/or but not both. Yes. You get your exact image with the prompt applied to it, unlike Gen-2 that just sort of uses it as a guide. I really suggest giving it a try; it's a fun toy and very neat, and the other players love it when you give them a Hogwart'd character pic, even if they can't actually put it on their VTT sheet. Megaman's Jockstrap fucked around with this message at 16:16 on Aug 9, 2023 |
# ? Aug 9, 2023 16:13 |
|
Megaman's Jockstrap posted:There's some pretty rudimentary AI video generation out there, and so in my quest to make my games as cinematic and luxurious as possible I decided to dive in. How do you attach an image in Pika. The instructions are dogshit
|
# ? Aug 9, 2023 17:21 |
|
type "/create " and then your prompt click on the "+1 more" click on the "image" that appears in the menu right above the chat box attach your image
|
# ? Aug 9, 2023 17:23 |
|
Megaman's Jockstrap posted:type "/create " and then your prompt
|
# ? Aug 9, 2023 17:52 |
|
|
# ? May 2, 2024 05:15 |
|
Fuzz posted:Just posting here so I can link said post in another thread and not get probated yet again: I'm about to post a new OP to refresh the thread, but Fuzz any chance can you send me working links from your post here? e to add: New thread here! Raenir Salazar fucked around with this message at 23:57 on Sep 6, 2023 |
# ? Sep 6, 2023 23:22 |