Type | |
Stats | 2,465 |
Reviews | (267) |
Published | Apr 5, 2023 |
Base Model | |
Hash | AutoV2 38737D8E91 |
A new build and training method for the SD 1.5 model.
This is a limited training preview of the upcoming 1.0 (in a few days 🤟 🥃)
![](https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/d59e3a58-237f-4451-68e0-ebcbb3414200/width=525/d59e3a58-237f-4451-68e0-ebcbb3414200.jpeg)
Some prompts can get you over-saturated so do encourage CLIP SKIP >= 2
All images used for training were generated prior by Latex.ID custom builds and then used for learning.
Full disclosure/training data/legals and the Latex DATASET (10k+ images) will be released for download including training info and prompts with ver 1.0 in the next few days.
This will be (most likely) the final build for 1.5.
Expect a better prompt control for SD 2.1 right after
![](https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c1ff91c6-d510-41c4-90ac-93954e382b00/width=525/c1ff91c6-d510-41c4-90ac-93954e382b00.jpeg)
Latex.ID - LoRa & LoCon - Latex Generator v0.2 available
Latex.ID - LoRa - Leather & Latex Leggings v0.3 available
Inspired by the toolkit of our partner SPYBG and the ToolKit for Digital Artists.
Here is our LATEX GENERATOR :P
The training was done on latex fashion clothing including corsets, harnesses, chokers etc.
All images were Masked trained to be separated from the background.
Faces intentionally are anime style for now and based on SpyBG tool kit and the official clean version of 1.5
TEXT2img
All text2img is generated on white background(accordingly to the clothing prompt).
Models can be extracted easily from that for easy usage for another purpose :P
Upscaling on those is decent for this release—no 4K yet(working on the next ver).
INPAINTING /img2img/
WORKS surprisingly REALLY well on this early version.
Just a quick sloppy mask and 0.60-0.65 denoising are needed for generating latex on your favorite person
All samplers are responding well. But to try DPM++ 2S a Karras for better and crisp results.
As for prompts/// training was done on Latex fashion and on a woman...
So try anything with latex and a woman
latex bodysuit outfit dress swimsuit corset collar outfit harness metal choker etc
colors of the latex CMYK :P
FULL PREVIEWS are available at:
https://latex.id/model-latex-id-cmyk-v06r/
Do share your toughs and if it's working for you.
This one was released way earlier. A lot more colors and different latex styles are coming next.
SAMPLE PROMPTS:
a woman in a red latex dress posing for a picture with her hands on her hips and a black collared collar around her neck, 1girl, metal chocker, harness
a portrait of a beautiful gorgeous woman in a latex outfit with harness and metal choker
a woman in a black and white latex posing with her hands on her hips while wearing a black and white corset top and white corset with a metal choker on neck, futuristic glossy+++ latex+ suit, render of a cute 3d anime girl, collar and leash, professional product photo, 1girl, solo, thighhighs, breasts, hands_on_hips
a woman in a red latex dress posing for a picture with her hands on her hips and a black collared collar around her neck, 1girl, breasts
Negative prompt: (((signature))), (((text))), (((watermarks))), deformed eyes, close up, ((disfigured)), ((bad art)), ((deformed)), ((extra limbs)), (((duplicate))), ((morbid)), ((mutilated)), out of frame, extra fingers, mutated hands, poorly drawn eyes, ((poorly drawn hands)), ((poorly drawn face)), (((mutation))), ((ugly)), blurry, ((bad anatomy)), (((bad proportions))), cloned face, body out of frame, out of frame, bad anatomy, gross proportions, (malformed limbs), ((missing arms)), ((missing legs)), (((extra arms))), (((extra legs))), (fused fingers), (too many fingers), (((long neck))), tiling, poorly drawn, mutated, cross-eye, canvas frame, frame, cartoon, 3d, weird colors, blurry
Steps: 40, Sampler: DPM++ 2S a Karras, CFG scale: 6.5, Seed: 3613218317, Size: 768×768, Model hash: 60231c9ba7, Model: CMYK.idle.latex.id-alpha6-R2, Denoising strength: 0.61, Hires upscale: 2, Hires upscaler: R-ESRGAN General 4xV3
PS: do experiment with clip skip 2 too
PPS: Model focus and training are is based on SFW custom datasets!
but in this early version, something might "pop up"