Hi, I’m a product and car designer, and I’m so excited to test with AI, I think is a good tool for designing. This tool is so useful for the design process (shapes-ideas generation), but more than that it helps so much to refine aesthetically. For enhance my trained models, I have also to mix and create different kinds of models, so I will publish some of these that I'm trying to collect.
FRESHPHOTO (realism eddiemauro-mix): Product of compiling several realistic models, it performs well in different types of photography
Recommendations:
It works from 512, but is also good with larger image sizes, from 768 and up. The 512 is much more creative at times, but it's not the rule.
Use DPM++ SDE Karras for more detail. Less detail: Euler a.
Use EasyNegative Embedding, it is mandatory.
Use high Sampling Steps, to generate better quality. Steps 30-50. The more steps, the image has more quality, 50 or more is the same. It is recommended between 40-50.
Use loras like EpiNoise and add_detail, and test. More than all the second. The add_detail generates much more detail, the results are interesting, it also increases the contrast. Add_detail is a Lora called Detail Tweaker, and you can download here:
https://civitai.com/models/58390/detail-tweaker-lora-lora
Use VAE 840000 to increase the contrast, but it works without VAE and gives a more natural touch. Others can be tried.
CFG Scale 7-9. Although, you can experiment. There are times when I notice that the CFG Scale 9 setup with Sample Steps 50 is good. More than 10 with high sampling steps start to fail.
To capture people's ethnic groups, it is recommended to use the name generation method, selecting the ethnic group and copying only a first and last name. This page:
https://www.behindthename.com/random/
Put in the negative prompt: ((asian)), Japanese, Korean, Chinese. Sometimes it has a strong effect on generating Asian faces. If this is the intention, remove them from the negative.
The photo style has a subtle hint of warmth (yellow) in the image. Also a soft touch of desaturation of the colors.
Hires.fix is not always useful in images, and it loses detail. Use img2img method, increasing the size of the image. Denoise from 0.1-0.2. Do not use "Restore Faces", it generally generates good eyes and faces, although it can be used but very subtly, ignoring its use generates more naturalness in the faces. Use Hires.fix experimentally.
Example Prompt:
Positive prompt: RAW photo of Gotzon Otxoa in casual clothes, little smile, small details, photorealistic, ultra-realistic photo, 8k uhd, dslr, soft lighting, high quality, film grain, Fujifilm XT3, (masterpiece) <lora:add_detail:1>
Negative prompt: ((nsfw)), ((asian)), Japanese, Korean, Chinese, deformed eyes, ((disfigured)), ((bad art)), ((deformed)), ((extra limbs)), (((duplicate))), ((morbid)), ((mutilated)), out of frame, extra fingers, mutated hands, poorly drawn eyes, ((poorly drawn hands)), ((poorly drawn face)), (((mutation))), ((ugly)), blurry, ((bad anatomy)), (((bad proportions))), cloned face, body out of frame, out of frame, bad anatomy, gross proportions, (malformed limbs), ((missing arms)), ((missing legs)), (((extra arms))), (((extra legs))), (fused fingers), (too many fingers), (((long neck))), tiling, poorly drawn, mutated, cross-eye, canvas frame, frame, cartoon, 3d, weird colors, blurry, (((many people))), multiple people, cgi, 3d, render, sketch, cartoon, drawing, anime, cropped, worst quality, EasyNegative
You can follow me on my social networks. I will show my process and also design tips and tools:
https://www.facebook.com/eddiemauro.design
https://www.instagram.com/eddiemauro.design/
https://www.linkedin.com/in/eddiemauro/
https://www.behance.net/eadesign1
