If looking for a image workflow

https://civarchive.com/models/526749/image-generation-workflow-up-to-full-subject-8k-images

this is my spiritual successor to this workflow

after 2 months they finally added this flow back to my profile so i can update.

V3 is a massive text to video / text to image workflow 550+ node monster.

Features... everything. Face, body enhance, frame interpolation, reactor face swaps, 2 models multi lora, animate diff, Ipadaptor, and damn near every node set i could figure out and integrate... It is truly a monster that even i have trouble following. Its not a workflow for novice IMO. But it offers a hell of a lot of options and is the best workflow if you can figure it out.

Offshoot is bringing it back to image gen. Its a faster version of V2 IMO and a few extra options.

Features - face/body enhance, separate eye color enhance, lower vram, 2 models, multi lora, and color control prompting.

If not doing video, V3 is overkill for most people.

Warning this workflow has over 550 nodes. Thus it is recommended to have 32gb system memory and catching buggs in this nest of nodes is going to be hard to pin things down so i may not be able to get it working for all.

Think I got most of the bugs, everything is working as expected for images solo or batched, videos working well, more options were added and new notes.

New V3 Animate diff models supported with the lora.

What this workflow does

This is an all in one workflow with text to image and text to video, multiple enhancements steps, face replacements, and more

New in this version

Full batch processing, hence allowing animate diff to work

Ip adaptor integration.

Cropping initial images

Full control net support

Reactor face swapping

Frame Interpolation

Many extra model options

LCM

2 masking options.

Control of masks used, no more background objects

Better Person separation.

Back and better

Face and body enhancing working for both batches of images and video through animate diff.  400 seconds 16frame 960x1440 video on my 3060 starting from 320x480

More efficient masking with less unneeded masking tasks. Might be faster not sure really.

2.0 release - The number of breasts and asses i have had to view to bug test this is absurd.

A 2.1 will come soon with a fixed UI

Short description of changes - many notes included in the flow.

Reworked from the ground up with learned knowledge

New body detection with a controlnet preprocessor that works almost perfect.

Batched or unbatched processing for lower vram cards.

Dual model, Dual lora loading, clip skip options, vae options,

Image to image repair/enhance existing images.

Optional Roop face replace/restore

Optional Controlnet tile upscaling.

Requires

WAS Node suite - for maths

https://civarchive.com/models/20793/was-node-suite-comfyui

Impact Pack - for SAM and pipe flow ease

https://civarchive.com/models/33192/comfyui-impact-pack

Masquerade Nodes - for many many mask tasks.

https://civarchive.com/models/52723/masquerade-nodes-comfyui

Controlnet preprocessors - for body masks.

https://github.com/Fannovel16/comfyui_controlnet_aux

Optional - if you don't want to use these, they can be deleted from the flow.

Control Net 1.1 - installed through comfy ui manager

https://civarchive.com/models/71980/comfyui-manager

Tile models - only need one

https://civarchive.com/models/38784/controlnet-11-models

https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/tree/main

Face restore nodes

https://civarchive.com/models/24690/comfyui-facerestore-node

Roop

https://civarchive.com/models/137245/roop

1.0 release

- 3 checkpoints with lora use now integrated one set for each area

- proper explanation of use

- final ui fixes.

V3 - final

3.0 fixes

- improved detection

- proper pasting on most images

- better blending

- added options

- better interface

- better support for lower base sizes

- faster generation

V2.5 working as intended... same thing just working now

Introducing the NSFW generator with HD face and now body enhancing. - This workflow can work some serious magic.

Requires -

https://civarchive.com/models/52723/masquerade-nodes-comfyui

https://civarchive.com/models/33192/comfyui-impact-pack

and

https://civarchive.com/models/20793/wass-comprehensive-node-suite-comfyui

Options not pulled out of nest yet

• mask to region (constraints and batch behaviour) depending on photo types can improve performance.

Beta 2

Changes - Everything... Big thanks to the adetailer, the masquerade team and the people that make the detection models.

- All faces now fix and enhance automatically. Bodies now do the same

- Can load either a xl or 1.5 model separately for all parts of The main scene, the bodies, and faces

- All changes have their own prompts and all respond really well.

- 100% modular, can bypass any parts you don't want. Just Ctrl bounding box the group and press Ctrl and b

- When doing photos with multiple people, all are enhanced, but each person takes more time. The more complex the photo the more trouble it may have.

- Any resolution is supported, can do some great things with Lower res bases and works quite well with 1.5 base resolutions.