I was using someone else's upscale LoRA. It worked well, but on female subjects it kept growing coarse, masculine body hair. I wanted an upscaler aimed at female portrait and fashion photography that restores real detail in skin, hair and fabric without drastic redrawing the person or hallucinating content.
The trigger is "zqvx, sharpen the image, enrich the original textures with more detail."
Use the included workflow(Absolutely required)
Download the json workflow from optional file, or google drive link:
https://drive.google.com/file/d/17u1cLp_P95FNpLIl41x342UQOoBHOtxq/view?usp=drive_link
Diffusion_models
Description
FAQ
Comments (5)
I feel like an idiot right now as I don't usually have too many issues with ComfyUI workflows. Trying to install, update, etc. to get the SeedVR2 node packs. What do I need to do to?
You can actually skip the SeedVR2 groups, since the first generation already gives pretty good results — SeedVR2 is just a second pass for extra enhancement.
@crossplayer2187189 Yeah I tried to bypass them so I deleted them but I was curious still to figure out what I am doing wrong to get them is all
Looking at the example images, skin still looks very much plastic, as with every other lora (probably Klein9b feature) :(
Yep. That's whole F.2K problem. Some creators here are making visual aesthetics LoRAs for both F.2K and Krea2. Dataset used is probably the same but output aesthetics are way different between these two models. Especially on LoRAs for retro aesthetics. Krea2 handles them just fine while F.2K puts up some resistance and applies it's native plasticky textures.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.