Custom LoRA High-End Photo Retouch AI Model

I want a reproducible AI model that can take my raw portraits and return a polished, high-end beauty retouch that mirrors my own editing style. Think pristine yet realistic skin, flawless blemish removal and nuanced tonal work—no plasticky faces or over-smoothed texture.

My current workflow lives in Photoshop and Capture One, but I’m open to any backend that gives the best fidelity; LoRA FLuX on a Stable Diffusion base has caught my eye, though I’m happy for you to suggest a stronger alternative if you can prove the advantage.

What I will hand over
• A curated set of before-and-after TIFFs/JPEGs—consistent lighting, color-managed, already retouched by me.


What I need back
• The trained model weights (.safetensors or compatible) ready to drop into Automatic1111 or ComfyUI.
• A concise README covering install steps, inference parameters, and recommended hardware.
• At least ten side-by-side samples using my unseen test shots that show accurate skin smoothing and blemish removal while preserving pore detail and minute tonal shifts.
• An optional script or notebook for batch processing would be a welcome bonus.

Acceptance criteria
1. Texture retention: pores and fine hairs stay intact at 100 % zoom.
2. Color neutrality: no unexpected shifts away from my reference retouches.
3. Instructions reproduce identical results on a second machine.

If you’ve built portrait-specific LoRAs or similar beauty pipelines before, send over a couple of uncompressed samples and a quick overview of your training stack so I can see how closely you hit this high-end finish.