[Leeloo Please JPG]
DO ME A FAVOR if you like what I'm doing...
Please consider to test my trained models with different topics and compare it with another models in order to see if my goodies are adding anything of value to the SD world. I will be glad to get a constructive feedback about what is GOOD and what is BAD about my models.
So, by adding the REVIEW with comments you will help me to get energy and motivation to move on.
Thanks in Advance!!!
!!!NotaBene!!! Load my example images to ComfyUI.
I use complex and simplistic ComfyUI workflows with upscaling and without. To get a tips for prompts you can press a "NN Nodes" button in image resources and paste it to ComfyUI.
Also, I often repeat some popular prompts from version to version in order to show the difference, so you can compare and get an idea of what's new.
MAIN GOAL
Main focus is a balance between photorealism, creativity, bokehlicious rendering with shallow depth of field, anatomy improvements, emotions, artistic and analog style.
PS: if you like what I'm doing you can Buy me a Coffee
Description
This version shows better creativity, color variations and vibrancy, human pose and creatures morphology variations, but slightly less stable. It renders good hands / fingers in many cases (in my way of prompting; load example imgs to ComfyUI). The chance of bad hands with this version is higher than with v1.2c, but noticeably lower than the same chance using JuggernautXL v5.
Positive Prompt example:
realistic full body shot of (elegant Slavic woman:1.15) with natural skin texture and intricate body art and perfect feminine anatomy showing open palm, abandoned industrial zone in Chernobyl, Voronoi-shaped blocks on the ground, dark carbon voxels with sharp edges, low-poly abandoned nuclear plant, (perfect hands:1.23), (glitter:0.85), (shiny black slime, Voronoi scales, depth of field, (swirly bokeh:1.275), (Kodak Portra 400:0.925) :1.25), (realistic complex abandoned nuclear plant background:1.1), shiny glass, (cosmic horror, gloom, light leaks, mysterious, surreal:1.1), remarkable masterpiece, celestial, ethereal, epic, magical light flares, natural soft dreamy lighting, ((warm-yellow and cold-green:0.85), cinematic look, ideal hand anatomy, hairy pubic, hairy armpits:1.25)Stay tuned for more examples posted. Read post test and read PNGInfo of my images for tips & trick of good prompting with CinEro merges.
Good Luck! And many thanks for feedback and sharing with other creative heads!
Will be appreciate if you Buy Me a Coffee ;)
FAQ
Comments (6)
does the model always generate a very similiar face, or is face and style generation more random ie is it overbaked or closer to sdxl freestyle
It shows variations. It understands difference between nationality (for example, Slavic vs Tatar). Also, it reacts on celebrity names without corresponding LoRa. For example, "[Audrey Hepbern : Scarlett Johansson : 0.777]" will make a good result. I will make some examples soon. Also, I want to try visual description of facial features. Stay tuned...
I have posted the example with another faces (didn't use celebrity names).
Because of the way these "Diffusion" model works, if you just specify say "woman" in a simple prompt, you will get a similar "average" face most of the time (the variation will come mostly through the seed). So as homoludens pointed out, to get more variations you need to be more specific in your prompt, by adding things such as race, nationality, age, hair and eye color, and the use of different names. Basically anything that move the prompt away from "the average".
@NowhereManGo I also noticed that adding some random number after the class of subject (like "woman 42") can also affect the morphology of the face, but I didn't properly check this guess...
@homoludens Interesting, but certainly possible, because if an image is tagged with the number 42 for some reason, then that image will add some variation.
But then all tokens affect the image in some ways, so maybe there is no "deep" reason of any kind here 😁
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.