If you like this then please buy me a coffee sometime: https://www.buymeacoffee.com/Darkseal

The images attached might look like a celebrity freak show for some but let me explain:

This is Pinocchio model with 1.5 face weights re-injected back into it using the merge block weights tool. Please see image below. By doing this we obtain the closest results to the famous actors and actresses 1.5 was trained on. In the celebrity images, I have not given much thought to the prompt, I have left face fix OFF, there is no hires up-scaling (except for the main image of Tom), there are NO embedding or loras. This will make using loras or embedding as close as possible at low power values while keeping the model as free as possible to prompt the environment. This way even the worst embedding\ti\lora trained on 1.5 should work as expected, let alone a good one.

Following the information found here: https://rentry.org/Merge_Block_Weight_-china-_v1_Beta#31

Please note... what we gain in faces we loose in penises. Yes, that's right. Gaining famous faces removed the heads of penises. When using penis prompt please use some kind of embedding or lora to correct the horror that is the penis heads in this model.

Show some love and buy me a cup of coffee above!

Support me on gumroad - https://darkseal.gumroad.com/

See more of my ai art on DeviantArt deviantart.com/darksealstudios

more 3d stuff - https://www.cgtrader.com/darkseal

more pron 3d for poser/daz for making your controlnet - https://www.renderotica.com/content-artists/Darkseal/201