A simple but powerful ComfyUI workflow that turns any 2D illustration, anime frame, comic panel or drawing into a realistic photo.
It uses Florence 2 to automatically caption the input image and convert it into a clean prompt, then generates a photoreal result with Flux 1 Dev + Flux Redux, a custom LoRA, Control Net for a better fidelity poses, Face Detailer for higher fidelity faces, and a light upscale with Ultimate SD Upscaler for extra sharpness.
Just drop your 2D image in, hit generate, and get a realistic version in a few steps.
This workflow includes a LoRA that is not created by me.
All credits for this LoRA belong to its original author:
👉 AIDigitalMediaAgency
📎 LoRA download: https://civarchive.com/models/726732/realistic-people-photograph-flux
I'm actually working on a v2 where no LoRA will be required at all.
Description
Features / Highlights
Works with anime, manga, comics, sketches and illustrations
Fully automatic: Florence 2 handles captioning and prompt creation
Powered by Flux 1 Dev + Flux Redux for strong realism
Includes LoRA integration, Face Detailer and Ultimate SD Upscaler
Step-by-step preview nodes so you can see the evolution from 2D to realistic
Modular: you can easily bypass or tweak each block (caption, LoRA, face detailer, upscale)
FAQ
Comments (5)
Can you guide on how to actually see the connections between these floating nodes i.e, setnode andgetnode. is there a simpleway to see it?
There isn't a simple way to see the connections with easy set and get nodes. The only way is check the names. model to model, clip to clip etc...
Thank you. I hope people stop using it then. In complex workflows, it makes it very hard to understand the connections and modify or tailor some to individual needs
@Umair11 I use it because it really bothers me to see cables everywhere, and I need to keep everything organized. But I understand that at first it might seem complicated for some.
"It uses Florence 2 to automatically caption the input image and convert it into a clean prompt" ... So it's not converting the actual image from anime to real just making a new image with the same visuals :/