Post B17SxFHdDhTKbEbjBA by CrazedWeevil@poa.st
 (DIR) More posts by CrazedWeevil@poa.st
 (DIR) Post #B17PTxnKrbgU6CmZQe by Kalogerosstilitis2RevengeoftheJunta@poa.st
       2025-12-10T21:53:39.852553Z
       
       2 likes, 0 repeats
       
       Local ai is officially no go. Ive tried pretty much every local thing comfy has in the templates. Everything downloaded. Either not enough vram (8gb) or it crashes or produces errors. I give up until they bring put a 5080 with more than 16gb
       
 (DIR) Post #B17Rh81YLHz7Uuq0Ey by CrazedWeevil@poa.st
       2025-12-10T22:15:37.072392Z
       
       0 likes, 0 repeats
       
       @Kalogerosstilitis2RevengeoftheJunta What model are you trying to use?
       
 (DIR) Post #B17Rh90Wgal2Y1MhHM by Kalogerosstilitis2RevengeoftheJunta@poa.st
       2025-12-10T22:18:26.196050Z
       
       0 likes, 0 repeats
       
       @CrazedWeevil All the local wan ones.
       
 (DIR) Post #B17SGRm0uGoYsuPW4G by Bakke@poa.st
       2025-12-10T22:24:50.361728Z
       
       1 likes, 0 repeats
       
       @Kalogerosstilitis2RevengeoftheJunta I don't think any of the templates will work with 8GB vram. You'll have to look into GGUF models.This should in theory help you get things running:https://www.youtube.com/watch?v=7hUO6KhUsvQ
       
 (DIR) Post #B17SxFHdDhTKbEbjBA by CrazedWeevil@poa.st
       2025-12-10T22:32:02.924387Z
       
       0 likes, 0 repeats
       
       @Kalogerosstilitis2RevengeoftheJunta It should be doable with 8GB is you use the Q3 or Q2 GGUF versions for WAN 2.2https://www.youtube.com/watch?v=7hUO6KhUsvQA fairly decent guide in this video, though I'd recommend trying the SmoothMix workflows...they're quite simple to use but still give you control over all the settings you'll need.https://civitai.com/models/1847730?modelVersionId=2264611How much system RAM do you have?
       
 (DIR) Post #B17SxGZkPpYcbi5i3U by Kalogerosstilitis2RevengeoftheJunta@poa.st
       2025-12-10T22:32:33.688566Z
       
       0 likes, 0 repeats
       
       @CrazedWeevil 32gb
       
 (DIR) Post #B17TGMHFRQGtnCXebI by CrazedWeevil@poa.st
       2025-12-10T22:35:44.073419Z
       
       2 likes, 0 repeats
       
       @Kalogerosstilitis2RevengeoftheJunta There are some workflows that have something called Block Swap, which allows you to use system RAM in place of VRAM for loading the models, helps with low VRAM GPUs. They are a lot slower though, but the price you have to pay I'm afraid.
       
 (DIR) Post #B17TMYEhtIPAcWJ2x6 by APPTeOORuzvlGOetVY.verita84@poster.place
       2025-12-10T22:37:06.140357Z
       
       0 likes, 0 repeats
       
       @Kalogerosstilitis2RevengeoftheJunta localAI was too unstable for me on different GPU's. Garbage. Are you wanting to do images or chat? YOu can do pretty good on chat with 8gb ollama + qwen
       
 (DIR) Post #B17ToMhMNjhAwcyGxs by Kalogerosstilitis2RevengeoftheJunta@poa.st
       2025-12-10T22:42:09.906822Z
       
       4 likes, 0 repeats
       
       @verita84 Im a simple man. I want a) memes from text and/or pictures modified with a text prompt. B) tits. Giant tits editor of pictures by text prompt. I wish i could make this like @Paleface did but i understand video is basically impossible in any meaningful timeline on my system. The memes are the real need for local ai. Shlomo will put me on a list gere in Europe if i start kike memeing on line. Zog doesn't care about tits, i could use an online one i guess but still if i can get the memes why not fo the tits as well you know?
       
 (DIR) Post #B17Tvc8VNnCGqq7n72 by APPTeOORuzvlGOetVY.verita84@poster.place
       2025-12-10T22:43:27.837026Z
       
       1 likes, 0 repeats
       
       @Kalogerosstilitis2RevengeoftheJunta @Paleface you can try using sdnext with the cli argument --low-vram or --med-vram to get more out of your 8GB
       
 (DIR) Post #B17UF4jPI5Ib2WfI48 by Bakke@poa.st
       2025-12-10T22:46:59.932753Z
       
       4 likes, 1 repeats
       
       @Kalogerosstilitis2RevengeoftheJunta @verita84 @Paleface Tits and memes are important.
       
 (DIR) Post #B17UTolouqJPsgp7iq by APPTeOORuzvlGOetVY.verita84@poster.place
       2025-12-10T22:49:30.113802Z
       
       1 likes, 0 repeats
       
       @posterchan @Bakke @Paleface @Kalogerosstilitis2RevengeoftheJunta ComyFUI has a low-vram mode!https://github.com/comfyanonymous/ComfyUI/discussions/1177once you get that running, you can run posterrchan to generate images on Fedihttps://git.poster.place/verita84/posterchangeni sexy jewish anime girl with small breasts. HIgh quality. anime ar  style.
       
 (DIR) Post #B18qbA0wiouvwdlUAa by zippy@poa.st
       2025-12-11T12:38:15.366310Z
       
       1 likes, 0 repeats
       
       @Kalogerosstilitis2RevengeoftheJunta I second trying ollama.  You should be able to get distilled models running.