MODEL IS GREAT, BUT IMAGES GENERATE HAS CHINESE OR JAPANESE FACES.
Thank you so much for this amazing fine-tuned model. I was waited for so much time.
Really appreciate your time and work into this.
I generated a lot of images, I found that with this model
at steps 4, face generated are based on prompts. Also, at this steps, there is lots of distortion.
but at 20 steps, they are almost in every image are Chinese or Japanese. Even explicitly mentioning the nationality of person.
The issue of plastic skin is still there for both jaguar and aesthetic model (I don't know is there any lora or CGF or guidance configuration that can solve this issue.)
Could you please provide me best configuration that can generate real person images without plastic skin. Is there any lora that can be used?
The 3.1 aesthetic model from you, is working very great even at 4 step. Jaguar model has so many defects at step 4.
At 20 steps this aesthetic model is giving faces of prompt nationality.
Also, is there any prompt issues? I am adding this in prompt "high quality, 8k resolution, ultra realistic, image captured on canon DSLR camera, photo, photography, real, 4k, detailed skin texture, face details, perfect face, perfect eyes, perfect hands, perfect feet, perfect anatomy,". and also adding this as a style :"Style Clean, futuristic minimalism with metallic finishes and stark lighting., fashion photography, glamorous, stylish attire, captivating pose, dynamic composition, high-fashion aesthetic, professional studio lighting, expert styling, showcasing trend, conveying elegance, conveying attitude, evoking emotion, creating visual impact, editorial storytelling, artistic interpretation, fashion-forward visual".
and in negative prompt :"text, watermark, logo, signature, jpeg artifacts, jpeg, ugly, low quality, low resolution, bad anatomy, three hands, three feet, six fingers, bad hands, bad feet, bad face, distortion, ".
Please guide for tips.
Thank you so much.