Finally v1.
I needed to see where the industry was going before I went with a version 1.
This uses Flux.1-Dev GGUF Q2.K Q3.KS Q4 and I did many tests. This can produce PRO levels of quality at half the time the Dev standard does. Yes I am using upscalers and segmenting with second pass. You can opt out of anything along the way. ALSO API support is attached if you want to use Pro directly with Fal.ai. loras also added.
Yes I removed controlnet and Ipdadapter. That needs work. I am sorry.
Please consider joining my Sponsor page so I can keep most of my work available for everyone - I'll be releasing early access models there for a cheaper monthly fee than buying them here individually, and also will be providing exclusive models and more!
For business inquiries, commercial licensing, custom models (LoRAs/checkpoints), and consultations, please get in touch at [email protected] or [email protected]. You can also contact me here through CivitAI DM or join my Discord.
v.07 has FLUX InPainting integration / Refiner and Upscaler added. (ControlNet has been removed until further notice)
Old article of initial workflow
Lut files also attached in v.07
Version 0.5 is FLUX ControlNet model attached. (I did not implement the LORA as I did not see significant results). Be aware the ControlNet is Canny and as of this time does have some bugs and I have been told there may be some changes coming soon.
[RESOURCES] Please see Notes inside workflow for Model URLS and info
v.06 has FLUX ControlNet integration updated.
v.05 has FLUX ControlNet integration built in and SDXL refiner.
========================================================
Version .03 is a work in progress for version 1 that will have Full Style transfer using Flux.
Basics:
You MUST update ComfyUI from Update folder (using both .bat files)
Then, open ComfyUI and Load the .json to open the workflow.
You will need Node Manager and use it to install some custom nodes.
https://github.com/ltdrdata/ComfyUI-Manager
Install:
If you get an error in any of the nodes above make sure the files are in the correct directories.
See the top of the examples page for the links : https://comfyanonymous.github.io/ComfyUI_examples/flux/
flux1-dev.sft goes in: ComfyUI/models/unet/
t5xxl_fp16.safetensors and clip_l.safetensors go in: ComfyUI/models/clip/
ae.sft goes in: ComfyUI/models/vae/
I got so much help from everyone here and on L2 & CIVITAI I thought I try to return the favor.
This Comfy workflow does require a few nodes besides the obvious FLUX by Black Forest Lab. I will assume you know how to install everything so this is not a beginners workflow.
Comfyroll Studio
rgthree
mixlab
ComfyUI WD Tagger
Florence-2
ComfyUI_LayerStyle
Pythongosssss
ComfyMath
and finally Image Chooser
You can always use Comfy Manager to install or disable them and run things directly.
Hope you enjoy and feel free to help add to the community.
Drag the workflow on top of your Comfy UI (L2 Image might work as well)
Use:
This was designed for you to be able to drop an image and Flux will just work without typing a ton of tags or descriptions.
Description
This Workflow has LORAs, Ipadapter and Controlnets by doing a second pass using SDXL
You do have the option to turn it off.
Some slight improvements on node setup
FAQ
Comments (9)
Guys have fun with this I am revamping it again and will have it up in a few days. Adding the Florence2 request and adjusting the nodes. Let me know if I should remove the SDXL parts.
Thank you.
I'd like to ask how things are going now and how the InPainting feature is supposed to work.
@Rance99 You pick inpainting add an image and mask the area you want to change. (right click on the image) then type what you want to change.
@zGenMedia # ComfyUI Error Report ## Error Details - Node Type: VAELoader - Exception Type: AttributeError - Exception Message: 'NoneType' object has no attribute 'lower'
I can't find lower or vaeloader to change this problem
@zGenMedia ok,i find it ,its behind load lora
@zGenMedia Now I try to inpaint, it just masks, but it doesn't work, it doesn't change my input image, it just outputs the masked image
@Rance99 Sorry for the late response. I will look into it now but I need more information. What comfy are you using (portable? Windows?) and are the other parts of the wf working for you? Thanks.
@zGenMedia win10
How do i enable trust_remote_code=true? I cannot use it as commandline arg. And i dont know where to put it in the model files.