/bbwai/ - bbwai

Stable Diffusion, Loras, Chatbots

BBW-Chan is supported by simple text or static image ads from a-ads.com.
Please consider whitelisting us in your adblocker if you enjoy your stay!


Mode: Reply
Name
Subject
Message

Max message length: 9999

Files

Max file size: 10.00 MB

Max files: 6

Captcha
E-mail
Password

(used to delete files and postings)

Misc

READ THE *RULES* BEFORE YOU POST!

[ / / ]

(249.81 KB 1216x1216 2.jpeg)
(215.00 KB 1216x1216 4.jpeg)
(217.97 KB 1216x1216 6.jpeg)
(221.62 KB 1216x1216 8.jpeg)
(218.04 KB 1216x1216 10.jpeg)
Sequence tool funnysocks 08/24/2025 (Sun) 20:31:25 No. 35477
I've been making a tool to help with making sequences and I wanted to get some input on what would be useful features. It's pretty much stuffer.ai but designed around parameter sweeps. The main issue for me is cost of GPU in the cloud so I've been thinking about crowdfunding it with a Patreon or whatever... What might be useful features? What might be useful features that are worth supporting the app to get? So far I have - Generated image caching (kind of essential for this kind of app) so you don't ever have to re-build the same image twice say if you undo - Sequence downloader - tick the images you want and get a zip with images numbered in the order you clicked them - Auto-generate (after delay) so you don't need to hit generate all the time and I was thinking of - (Paid for) attach your own GPU(s) as a private worker instead of using the pool - "Cross fade" of tags, kind of like what you can do with the 2D grid right now but less wasteful on images
Maybe a feature to make videos?
I would pay a good amount of money for this IF I could run it locally like Comfy or Automatic. Looks like a pretty genius design tbh
>>35483 You can do this in a1111 with some loras by varying their strengths using the advanced tech of pressing insert and changing up numbers, no need for a tool https://www.deviantart.com/5arm/art/LORA-failure-1121279035 Ofc having a GPU that can run bigger models is bad. But there's going to be $3k dollar machines using AMD's CPU that can load up to 128 gb models and run them at a reasonable pace at home. (not sure how fast but it's not slow)
(81.26 KB 894x894 1.jpg-pre.jpg)
(79.08 KB 894x894 2.jpg-pre.jpg)
(77.48 KB 894x894 3.jpg-pre.jpg)
(72.79 KB 894x894 4.jpg-pre.jpg)
(71.73 KB 894x894 5.jpg-pre.jpg)
>>35486 But it's a pain in the ass to iterate if you're always doing change > generate > change. Even worse if you want to go back to something you did earlier. The point of this is to make it easier and more fun. e.g. it allows you change a component of your main prompt and see that change reflected across your whole sequence. It really does save a bunch of time. Anyways here another I whipped up in a few seconds just because
(74.90 KB 894x894 6.jpg-pre.jpg)
(75.86 KB 894x894 7.jpg-pre.jpg)
(79.80 KB 894x894 8.jpg-pre.jpg)
(85.58 KB 894x894 9.jpg-pre.jpg)
(83.00 KB 894x894 10.jpg-pre.jpg)
Got a few more on my DeviantArt too https://www.deviantart.com/funnysocks55
Id be interested, would love to try it out. Currently doing something like this manually and it’s annoying.
Why not just help out stuffer.ai with your code idea instead of forking it? They need help with gpu funding too and it would be a much better use of resources if there was a collaboration
>>35561 It's not really a fork is it, unless you mean conceptually. And even then there are some things I'm less interested in doing like the image rating stuff (or really anything that requires image persistence because that's way closer to legal grey areas than I want to get) I mean there are several good reasons to and not to. For - It's a great project already (it was clearly the inspiration) and giving back would be really cool - It's decent funding already and cloud GPUs are spenny AF But also - I haven't run my own thing like this before so I want that experience from scratch - I want to offer an offline version too - And maybe this has applications outside of fetish porn too so I'd maybe like to explore that I've got some ideas on how to licence to run self hosted deployments so if I ever developed enough features for it to be useful as an alt frontend for all that existing GPU grunt I'd definitely consider contributing it's use
I've made a very rough demo available at https://nsfw.promptexplorer.io/ It only has epiCRealism v17 + DMD2 available and I've deliberately gimped the capabilities to keep image generation fast enough on the limited resource I have to show off the tool. There's also some bugs with sequence download at the moment but you can rename the unzipped files to .avif to get them to work :)
>>35632 Good stuff, can't wait for future progress
(71.94 KB 512x832 1.jpeg)
(69.95 KB 512x832 2.jpeg)
(70.44 KB 512x832 3.jpeg)
(69.42 KB 512x832 4.jpeg)
(66.62 KB 512x832 5.jpeg)
(64.92 KB 512x832 6.jpeg)
I've added a share button to create links that can be followed to load up a prompt now e.g. https://nsfw.promptexplorer.io/create?wf=epicrealismxl-v17-dmd2&p=20+year+old+white+college+girl+is+sitting+in+gaming+chair%0A%0A%28topless%2C+%28%28pink+panties%29%29%29%2C+side+view%0A%0A%28perfect+thinner+youthful+supermodel+face%29%2C+%28e-girl+makeup%2C+heavy+eyeliner%29%0A%0Arealistic%2C+vibrant+dark+colors%2C+%28long+pink+hair%29%2C+hands+at+sides&w=512&h=832&n=4&g=1&sd=1906159794&smp=lcm&sch=exponential&xt=obese&xw=1&xi=0.1&xc=5&yt=skinny&yw=1.3&yi=-0.6&yc=2 will create these images! I've also got custom LoRA working locally but there' still a bunch of work to get that into production because I need proper database and object storage first...

Delete
Report