Support my work-> SFW:
https://www.patreon.com/c/AIKnowledgeCentral
↘️ If you want to support my work NSFW: ↙️
https://patreon.com/FetishForge
I used this tutorial to create the LoRA:
This is my first flux lora so be gentle.
I you want to support me you can buy me a coffee:
https://ko-fi.com/aiknowledgecentral
or the visit my SFW Patreon page: https://www.patreon.com/AIKnowledgeCentral
2 / 10 cases you will just get a person with a ball in her mouth and sometimes the ball will be quite big, so i will improve on it in the future.
use prompts like:
cinematic erotic full body photo of a woman in black latex, with blue ball gag in mouth,kneeling on a sofa,leather straps, 4K, hdr,
I use the basic flux schnell workflow and a normal lora loader like you would use on SDXL.
If you support me with Buzz i can make more.
Description
FAQ
Comments (12)
Dude, this is expensive to use! The minimum generation is like 30 buzz. I am worried if I have to spend 4 on Kerras instead of 2 or 3 on Eular A. I guess for model maker, it's chump change since you guys get a lot buzz for people using your models, but the non-model maker users barely get any Buzz at all. I tried making a few Pony models on Civit Training, but they were shit and I spent a lot of time curating images and wasted Buzz making them. Can't wait to upgrade my computer...
Jepp it is expensive, i don't make the prices...
@AI_After_Dark not blaming anyone here, just saying, I can't imagine Flux gaining much traction.
@ponyboi after the disappointment with SD3, Flux has gained much traction already unfortunately. I still hope Stabillity AI plays their cards right with SD 3.1 because SD3 is much less hardware hungry.
@AI_After_Dark Interesting, I've largely ignored Flux. I'll have to become more acquainted with why people bother with it, but still, the generation cost is too high for me. 😓
@ponyboi That is very expensive indeed. Have you tried to run this locally? It is a lot of work, but the reward is astonishing
@sczakana No, I don't have a compatible GPU yet. I am waiting to upgrade my crusty old Intel Mac to latest M4 MBP in November. Already considered building a headless PC render box that just fields API calls, but I would rather plow that money into latest Mac chip. It's supposed to be epic. Also considering Digital Ocean PaperSpace since I am a software engineer and like to build SaaS tools. I want to build a private self-hosted image generation and model training tool that is as easy to setup as WordPress has been. As an app developer, I find Python workbooks to be a joke, the online image generators to be way too limiting and render box at home way too costly.
Got much better consistency moving the lora tag to the front of the prompt instead of at the end. fp8 dev on Forge v2
Any thought about a spreader or dental gag?
Currently working on a Hogtie + Ball gag, Hogtie + Ring gag lora, Spreader, Spider etc. is on my list
@AI_After_Dark Why is the flux-dev-vae used instead of ae.safetensors. Is this an experimental choice?
I never knew clip_l_sdxl_base would work. I will try that.
The clip l and the t5xxl encoders are the same as for SD3, i just did not rename them. The vae however i did rename from ae.safetensors to flux-dev-vae, that's because i also use the schnell model from time to time and did not want to mix them up.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.