|
moist banana bread posted:There pretty much is with just a few caveats. At least for a computer nerd it's there. Please do try. Finally got around to trying this, and either I'm doing something wrong (which is extremely likely) or my laptop is just not built for ai gen stuff (which is also extremely likely), but I'm having nil luck with it. I keep getting out of memory errors. Still, it wants to work, which is way more than I had before, so that was nice.
|
# ? Feb 6, 2024 06:10 |
|
|
# ? May 28, 2024 14:42 |
|
Stupid_Sexy_Flander posted:Finally got around to trying this, and either I'm doing something wrong (which is extremely likely) or my laptop is just not built for ai gen stuff (which is also extremely likely), but I'm having nil luck with it. I keep getting out of memory errors. SDXL requires significantly more resources to run than any of the SD 1.5 models do. I can't help you further without knowing what kind of laptop you have. Specifically if you have a video card on it and how much VRAM it has. Stable Diffusion can be made to run on CPU only, however, expect several minutes for each picture generated.
|
# ? Feb 6, 2024 06:27 |
|
Oh yeah the controlnet flows use 10GB of ram. I have had success using system ram in Automatic1111 but it slows generation times too much to be artistically pragmatic. You could try the v1-5 model to see if it's working. (edit: cough, this model) set resolutions to 512x512 moist banana bread fucked around with this message at 07:13 on Feb 6, 2024 |
# ? Feb 6, 2024 06:32 |
|
At the risk of showing my lack of programming knowledge, at what step in the above mentioned list would I change to that particular set?
|
# ? Feb 6, 2024 07:50 |
|
Stupid_Sexy_Flander posted:At the risk of showing my lack of programming knowledge, at what step in the above mentioned list would I change to that particular set? If you already have AUTOMATIC1111 installed then it's a simple matter of putting the models in the right folder for them to appears as options to change to when you have the webui itself open. In the base stable-diffusion-webui install folder you open /model/Stable-Diffusion and put however many SD models you want in there. The webui will run both SDXL and SD 1.5 types of models with no issue. When running, there's a drop down box in the top left that lets you choose which one is current loaded and being used. This lets you try out multiple models, like a graphics or cartoon or realistic model and just change them at will. If you want to try different model types out http://civitai.com/ is the main site for more. There is a porn filter on that site, feel free to use it. Let me know what kind of graphics card your laptop has because there might be another step required to run this. Editing the webui-user.bat file itself to add command line instructions is not hard but I can't know which command lines would help you without knowing your graphics card.
|
# ? Feb 6, 2024 11:52 |
|
After checking with the website, apparently my video card on my laptop is : NVIDIA(R) GeForce RTX(TM) 3050 Ti 4GB GDDR6 It's not a very high end laptop. It's a dell g15 5510. I'll do through and try to figure out how to dload the new model sets. Thanks for the help!
|
# ? Feb 6, 2024 14:40 |
|
Stupid_Sexy_Flander posted:After checking with the website, apparently my video card on my laptop is : 4GB VRAM is going to be tough I think SD 1.5 with --lowvram (might be double dash lovram) is as good as you'll get. I'm pretty sure 4GB is workable now with Xformers, I know 8GB was considered the lowest a few years ago, then 6GB was workable then became stable, and I'm pretty sure 4GB has become viable but you're going to be limited to pretty basic usage. To the point it might be better sticking with a cloud solution. You can run A1111 in google colab which is free, Google will give you like 4-12 hours every day on their hardware to gently caress around with this stuff. I'm not sure what the current recommended package is but you can very much run Stable Diffusion on Google's hardware free. That said there is something about running it local and being able to just do a batch of 100 images and have it run overnight or whatever. It's fun to see the proof and render offline.
|
# ? Feb 6, 2024 14:56 |
|
Thanks bing! anthropomorphic gingerroot with googlyeyes sitting at a table. He is drinking coffee calmly saying "this is fine". The room is on fire, the flames lick up the walls and coat the floor or a silver tabby cat e: loving around with Google abuse some more - I don't think any other current image gen could output images of a scene this complex with a Trump this Trump looking. pixaal fucked around with this message at 18:19 on Feb 6, 2024 |
# ? Feb 6, 2024 16:34 |
|
Stupid_Sexy_Flander posted:After checking with the website, apparently my video card on my laptop is : There is your issue. 4GB VRAM is not a lot when running things locally. It can still be done but you have to make tradeoffs in terms of speed when generating. In your base stable-diffusion-webui folder, right click the webui-user.bat file and Edit. This will open it in notepad so you can change some options. Change the "set COMMANDLINE_ARGS=" line to "set COMMANDLINE_ARGS=--medvram --xformers" and try to generate an image. If it works then you're good to go. If you still have issues try changing the --medvram to --lowvram and that should do it but very slowly. Xformers is an optimization program that will be automatically downloaded when you run. It's highly recommended. If you have any issues after that let us know and we'll try to help you from there.
|
# ? Feb 6, 2024 21:33 |
|
Yeah I basically heard gamers going off about how terrible the 4060 16GB is, realized how it's actually super duper neato, it made my computer quieter and better in every way. Also playing games is much more fun now and no a used 3000 series card would not have been a better value because I'm 30 now and give a poo poo about the part where I don't need a little 3d printed column just to support the dang thing or itll rip my pci slot out and also there's way less chance is roasts my 10+ year old psu.
|
# ? Feb 7, 2024 02:58 |
|
folks, it's text-to-3d time https://app.meshy.ai/workspace/text-to-3d it's obviously based on SD1.5 and the results are janky, but it's a sign of things to come
|
# ? Feb 7, 2024 07:45 |
|
Great, an additional dimension for us to struggle with
|
# ? Feb 7, 2024 12:46 |
|
null_pointer posted:Great, an additional dimension for us to struggle with 2D hands are bad enough. Can you even imagine the horrors of 3D AI hands?
|
# ? Feb 7, 2024 12:51 |
|
A Calabi-Yau manifold of endless mis-shapen boobs and dongs Edit: gently caress, someone's actually going to make that then jerk off to it, aren't they
|
# ? Feb 7, 2024 12:55 |
|
LordAdakos posted:2D hands are bad enough. The Good The Bad The Ugly I thought City Skylines 2 used AI generated models and that's why the performance is so terrible. I don't know much about 3D models but this feels really lovely looking for 10,000 polygons. quick google I'm seeing full character models listed as 9000 or so poly (face = polygon right?) at similar overall quality to the hand so you know just cram a full character model worth of processing into just the hand. pixaal fucked around with this message at 14:08 on Feb 7, 2024 |
# ? Feb 7, 2024 14:02 |
|
I used one of my own photographs for the --sref parameter with Midjourney and got some fun results. My photo: AI stuff:
|
# ? Feb 7, 2024 16:24 |
|
Google being a pain about words still somehow managed to pull this off reference https://www.youtube.com/watch?v=bPqwy6gv0KQ
|
# ? Feb 7, 2024 17:10 |
|
Pixaal, that was... Chilling.
|
# ? Feb 7, 2024 17:22 |
|
ScreenDoorThrillr posted:folks, it's text-to-3d time fondant.design doing the same thing with fairly similar looking results too
|
# ? Feb 7, 2024 17:30 |
|
Wow am I bad at prompts.
|
# ? Feb 7, 2024 18:12 |
|
Stupid_Sexy_Flander posted:Wow am I bad at prompts. a) what tool are you using (extremely important) 2) what are you trying to make, this thread always loves A Challenge
|
# ? Feb 7, 2024 21:15 |
|
I was checking out 3d one linked one above. Wasn't trying to do anything in particular, just threw a couple prompts into the creator that I thought would work well and it turns out nope, no they won't. I just gotta learn to be way more specific I guess.
|
# ? Feb 7, 2024 21:39 |
|
AI Art: I thought it would work well and it turns out nope, no it won't.
|
# ? Feb 7, 2024 21:46 |
|
|
# ? Feb 7, 2024 23:19 |
|
I made a website that generates hedgehogs. So I had some family stuff happen (spoilered for sad): Found out my mother definitely has at least one kind of cancer, and maybe two. Still waiting on the oncology results. She loves hedgehogs. My mom loves hedgehogs. Whenever my mom visits, she has me generate some. Like hedgehog knights, snowmen, ballerinas, etc. So I set up a system to make a new one each day. It's split into two servers. The first is in my basement and runs ComfyUI on my old 3060 12GB. It generates a new picture in the morning and uploads it a second server, which is just a cheap, single-CPU VM in the cloud. It's all pretty basic, but you can check it out at https://thedailyhedge.com/. The prompts are a combination of stuff me and my friends came up with, plus a bunch of the styles that included with Fooocus. It's been running since Tuesday, and each day I get a text message from Mom talking about it, so I guess I give AI images a big thumbs up.
|
# ? Feb 10, 2024 00:10 |
|
sending good hogs and good vibrations
|
# ? Feb 10, 2024 03:17 |
|
I'm sorry about your mom here's some hogs for good spirit milkandbananas fucked around with this message at 04:21 on Feb 10, 2024 |
# ? Feb 10, 2024 03:50 |
|
Definitely going to save those and show my mom , thanks everyone!
|
# ? Feb 10, 2024 05:04 |
|
trying out google
|
# ? Feb 10, 2024 05:12 |
|
repugnant posted:My mom loves hedgehogs. Sending hogs with best wishes! small fluffy cartoon hedgehog, on stage, buddy holly glasses, microphone Steps: 48, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 1, Size: 512x512, Model hash: f65275abcf, Model: notreallyistic_v2, Denoising strength: 0.32, Clip skip: 2, RNG: CPU, Hires upscale: 2, Hires steps: 32, Hires upscaler: 4x-UltraSharp, Version: v1.5.2 adorable cartoon hedgehogs, playfully romping through field of wildflowers, rainbow landscape, sparkling like diamonds in the sunlight Steps: 48, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 1, Size: 512x512, Model hash: f65275abcf, Model: notreallyistic_v2, Denoising strength: 0.32, Clip skip: 2, RNG: CPU, Hires upscale: 2, Hires steps: 32, Hires upscaler: 4x-UltraSharp, Version: v1.5.2 adorable cartoon hedgehog, eating giant watermelon, midsummer picnic feast, not a care in the world Steps: 48, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 1, Size: 512x512, Model hash: f65275abcf, Model: notreallyistic_v2, Denoising strength: 0.32, Clip skip: 2, RNG: CPU, Hires upscale: 2, Hires steps: 32, Hires upscaler: 4x-UltraSharp, Version: v1.5.2 1980s Heroic Hedgehog comic book cover, flat printed animation, Adventure Mouse crossover, exciting dynamic duo action pose Steps: 48, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 1, Size: 512x512, Model hash: f65275abcf, Model: notreallyistic_v2, Denoising strength: 0.32, Clip skip: 2, RNG: CPU, Hires upscale: 2, Hires steps: 32, Hires upscaler: 4x-UltraSharp, Version: v1.5.2
|
# ? Feb 10, 2024 05:21 |
|
Tapdancing Hedgehog from MJ
|
# ? Feb 10, 2024 06:39 |
|
|
# ? Feb 10, 2024 09:29 |
|
GPT4+Dalle3 snow angel skypirate
|
# ? Feb 10, 2024 14:12 |
|
Well I don't my mom's ummm... tastes? But I guess I'll save them all.
|
# ? Feb 10, 2024 14:14 |
|
bing prompt: hedgehogs sitting around a coffee table, there's a ziplock bag full of dried oregeno on the table. the hedgehogs have lighters in their hands. bing prompt: hedgehog with big muscles standing on the beach, selling dried oregeno in ziplock bags, the hedgehog has dreadlocks. LifeSunDeath fucked around with this message at 19:13 on Feb 10, 2024 |
# ? Feb 10, 2024 19:10 |
|
I adore the Pirate-hog
|
# ? Feb 10, 2024 19:13 |
|
Have an assortment of things. Some eldritch/horror stuff: Stranded on another world: God drat it I tripped again: And of course some hogs:
|
# ? Feb 11, 2024 02:06 |
Punk Hogs rock out.
|
|
# ? Feb 11, 2024 03:32 |
|
Icedude posted:And of course some hogs: Those are great! Would you mind posting the prompts?
|
# ? Feb 11, 2024 03:48 |
|
|
# ? May 28, 2024 14:42 |
|
made with the burger king whopper generator
|
# ? Feb 11, 2024 03:56 |