This series is focused on women and the female body, in every kind of setting or style. It can do fine art, social media style, surrealism, cinematic, and pornographic very well. The main goal is to have as much diversity as possible in an SDXL checkpoint. Diversity in ethnicity, body type, faces, posing, personality, and locations.
If you like it post your creations!
Donations to my Whop account will exclusively go towards hardware upgrades. Civitai Buzz will be used for site generation and promotions.
Upgrading my local hardware is my on-going goal so I can train locally.
Local training is imperative today. With Ai data centers there will be a vast drain on resources in the areas they operate. Since model & lora training will no doubt be hosted in the new centers, I want to be able to run locally only. The moment my Ai hobby causes actual harm, on any scale, is when I start collecting stamps instead.
General Prompt Guide
Weights in my prompts are primarily for testing the model's capabilities and trying to uncover limitations during test generation. However, I've created the tutorial below to help understand how model weights can be used more effectively.
Positive Attention
Weights between 0.5 & 2.0 are usually sufficient.
I sometimes use weights between 2.0 & 3.0 when I want to force attention to a specific ethnicity for testing.
Higher weights are also helpful when you want to magnify size or specific details. The difference between: large breasts, (large breasts:1.2), & (large breasts:1.8) is obvious.
Recognized weights are usually -5.0 to 5.0 with Dynamic Prompts.
Attention Equation
The parentheses ( ) on their own adds attention weight of 1.1. Unless you specify a weight.
(hyper-realistic) = 1.1
((hyper-realistic)) = 1.21
((hyper-realistic:1.3)) = 1.43
Brackets [ ] are negative attention weight of 0.91, not negative weight, it's just attention that starts under the 1.0 default of all words without weights. Stacking them decreases that number.
[deep shadows] = 0.91
[[deep shadows]] = 0.81
[[[deep shadows]]] = 0.72
Negative Attention
Weights over 3.5 are not usually effective or are a detriment during tokenization.
Negative weights do not have much effect in prompts.
Using too many high weights in the same prompt can cause distortions & deformities.
Using high weights will give specific keywords more attention, but that is often at the expense of other keywords.
Bonus
Prompt with high weights:
(HDR:2.2), cinematic, (natural sunlight:1.23), (hyper-realistic:1.45), (full body shot:0.6), masterpiece, incredible details, [feeling of solitude], mysterious, (deep shadows:0.72), Beautiful (30yo Malaysian woman:2.52), hazel hair, (mid_length grungy hairstyle :1.43), (natural large breasts:1.6), (wearing an unbuttoned stylish silk shirt), (brown pearl g-string, leather lace sandals:1.41), (sexy pose, reclining slightly on elbow:1.68) on a stunning designer wicker couch, at a rustic beach house rooftop with driftwood furniture, seashell decor, a view of the ocean, perfect for casual summer parties, sensual atmosphere, beautiful, guilty pleasure, shot on Canon EOS 5D, RAW.
Prompt without high weights:
(HDR:0.86), cinematic, (natural sunlight:0.8), (hyper-realistic), (full body shot:0.6), masterpiece, incredible details, feeling of solitude, mysterious, (deep shadows:0.72), Beautiful (30yo Malaysian woman:1.32), hazel hair, (mid_length grungy hairstyle), (natural large breasts), (wearing an unbuttoned stylish silk shirt), (brown pearl g-string, leather lace sandals), (sexy pose, reclining slightly on elbow:1.12) on a stunning designer wicker couch, at a rustic beach house rooftop with driftwood furniture, seashell decor, a view of the ocean, perfect for casual summer parties, sensual atmosphere, beautiful, guilty pleasure, shot on Canon EOS 5D, RAW.
To force clip & vae in FP32
Forge:
./webui.sh --clip-in-fp32 --vae-in-fp32 --no-half-vae(or add the commands to your webui-user.bat (Windows) or webui-user.sh (Linux) in
COMMANDLINE_ARGS= --clip-in-fp32 --vae-in-fp32 --no-half-vaeFor all commands (many are obsolete now since the developers have made them default)
Comfy:
python main.py --fp32-text-enc --fp32-vaeI don't use Comfy, so this is what I've seen on the web. Here is a link to all commands:
https://gist.github.com/ravetank/cedcd1ea89738b2e79b9f01e45a1a03d#file-comfyuihelp-txt
Description
This is basically a re-upload of v7.0 fp16, but the internal clip encoders have been replaced with:
LongCLIP Distilled fp32 Clip L & Clip G pre-trained
The new clips have better training.
Details:
Adds more overall image detail
Better prompt adherence
A lot more accurate when tokenizing prompts
Better prompt refining based on context
Just better gens, period
Base model is the fp16 precision version. The bigger size is because LongCLIP G is 2.8 GB, and the original clip is 1.6 GB.
Felldude Links:
Long Clip:
https://civitai.com/models/1805024/long-clip-distilled
You may need to force clip fp32 in Forge & Comfy.
--clip-in-fp32
FAQ
Comments (9)
This is a brief review of the Reality Bound XL model. I mainly test realism models for SDXL, Pony, and Illustrious. Among the models I've tested, Reality Bound was the most realistic SDXL model that produces the best images. In particular, the results are the best when applying LoRAs based on BigASP as the base model. I like that it reproduces East Asians well.
I also tested each version, and here are my personal impressions by version. v2 has a color tone like the movie Matrix. v3 does not produce clean images when applying LoRAs. v4 and v5 are similar but feel slightly different. Still, v5 seems to produce somewhat cleaner images. v6 and v7 did not feel better.
So my conclusion is that v5 is the best version of this model. Of course, this is my personal opinion, so please take it as reference only.
Helpful actually. Thanks for taking the time to benchmark. I will keep this in mind while I work on the next version. Oddly, v5 is a mix of the original 16GB that made v1 & 1.1, and 2 unpublished versions made prior to v4.
Are people getting different images out of the gutpunchv7 vs v71Fp32? When I do same seed comparisons with my external workflow I get the exact same image.
This is an excellent question and I hope someone can give you a real answer to it. Some info that may be relevant:
I uploaded 7.1 as a test. It is the same checkpoint as 7.0, just with the clips swapped. I would assume the same seeds would produce the same images. The clips are fp32, but I think Comfy and Forge will use them at fp16 unless you force fp32 in the config. I don't use Comfy, so I don't know for sure. In Forge, no workflows, I see a noticeable improvement in details and prompt adherence. But I admit that I very rarely test seed reproducibility.
@ironically_irascible do you have the cmd argument (the one in webui-user.bat) for forcing fp32 in forge?
@ShencilLizard I run on Linux, but the format for the .bat config should be:
set COMMANDLINE_ARGS= --no-half-vae --clip-in-fp32 --disable-gpu-warning
Good one. Thank you for this model and making it free to use.
fp32 v 16: will I see better prompting with 32vs16?
In theory yes, according to testing done by other people with fp16 vs fp32 clips. My experience here is that clips with different training will produce more details and will have better training on a wider variety of concepts. It honestly comes down to preference. They both seem to have a better grasp on anatomy in general, though.
The fp32 definitely adds almost imperceptible details that that fp16 doesn't. However, both clip versions were also trained by entirely different people. So you're going to see variety in prompt handling with either v8 model.
Felldude has info on this too: https://civitai.com/models/1044804/clip-l-and-clip-g-full-fp32-zer0int-and-simulacrum
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.