Frank Frazetta (1928–2010) was an iconic American artist, often hailed as the "Godfather of Fantasy Art." Renowned for his dynamic and dramatic style, he created vivid illustrations for comic books, paperback covers, and movie posters, particularly in the fantasy and science fiction genres. His work on characters like Conan the Barbarian and Tarzan left a lasting legacy, influencing countless artists in the field.
NOTE : Base model : Z Image Turbo & Flux1.
Showcase images were created with ComfyUI and are in PNG, include metadata & workflow.
Description
Model trained on Civitai with 44 images
Recommended resources : Fluxmania IV
Recommended settings : dpmpp_2m / sgm_uniform / 25 steps / guidance : 3.5 to 4.5
Weighting : 0.4 to 0.6
FAQ
Comments (6)
Sadly the Flux torso looks worse than ever in these renders.
Thanks for Uploading this awesome lora Adel_AI, Hey by any chance have you got any links to the t5 fp8 and the other text encoder you are using "Vit-L-14.....) model i cant find the exact one you are using! thanks
Thank you for your feedback.
The T5 fp8 is the standard model published by BFL
For the Clip L fp32 here is the link: https://civitai.com/models/1044804/clip-l-full-fp32-zer0int-and-simv4
@Adel_AI thankyou appreciate it
Love your loras! Thanks for sharing them! I'm new to training loras for flux and am using kohya. What settings do you recommend for training a style using 19 images that are 1024 in size?
Thank you for your feedback.
I can't answer your question, it doesn't only depend on the number of photos in the training base, it's linked to many other factors (image style and their convergence, caption quality, very stylized graphic style or not ...) which are as determining as the training parameters. I made about sixty models all trained on civitai, I always had a different approach and it happened to me very often to resume training 2 or 3 times to make a change.