V17 Update - Relight + Animation Enhancer (click show more / scroll down)
V15 Update - Ip Adapter Support !
V14 Update - Training Tools, helps to find the best intermediate models.
V12 Update - Flux Live Portrait, Mimic Motion & Flux Video Detailer
V11 Update - Flux Control Net (Depth, Canny, HED) plus Controlnet Inpainting with Lora.
V10 Update - Flux Inpainting with SAM2
V9 Update - Scene Creator - (click show more / scroll down)
V8 Update - *two videos* - (click show more / scroll down)
V7 Update - (click show more / scroll down)
V6 Update -
V5 Update -
V4 Update - (click show more / scroll down)
V3 Update -
V2 Update -
V1 Update - (installation & basic version)
Foda FLUX.1 Pack for ComfyUI
article (pack v3): New Emergent Abilities of FLUX.1 Schnell [ image 2 image ]
v1
Basic
txt2img / img2img mode switch
8K native tiled upscaler
v2
Flux Shift
allows model shift to be controlled
Flux Sampler Garden
all samplers preset for img2img, denoise variable
v3
Flux Shift Line
No Easy use version of v2 Flux Shift (lines only)
Flux Auto Shift
Magic prompt + one button prompt nodes to enhance prompts automatically
Flux Auto Shift Line
No Easy use version of v3 Flux Auto Shift (lines only)
v4
updated to all previous workflows, using low sigma shift and aspect sized dimensions.
added new modules for Color correction & RemBG
SDXL image Bridge, allows use of SDXL checkpoints,lora,cnets
.cube LUTs free downloads added to the bottom of this page
v5
Flux Bridge Kolors
Kolors txt2img/img2img with CNET/IPA
txt2img/img2img with Flux1
BOOSTED Flux Upscaling ! (new feature)
all previous modules included
V6
Flux Kolors Bridge ZENKAI
my own custom prompt styling system for flux bridge
Pure Flux ZENKAI
only Flux txt2img/img2img using my Zenkai prompt system
Flux Boosted Upscaling
image loader input, dedicated upscaling
fastest 2X upscaler on Flux
can run at 0.01 denoise with good results
BOOST YOUR SHIFT !
V7
Foda Flux Thumbnails
Has one job only. Make Thumbnails with AI Text.
three separate controllable stages, text, background, subject
many masking options for composition
create thumbnails or posters
full control over each layer
fast and clean
V8
Foda Boosted Upscaler
QoL improvements to the v6 release
Floda-Flux-Lora
Lora Stack Loader for Flux Workflow
Foda Webcam
fun experiment with webcam into Flux
V9
All 8 Main Workflows refactored with a full set of latest versions, using starndard connecting lines only.
due to popular demand with EasyUse not working for everyone :) Big Love
Foda Scene Creator
generate layers with Flux, using prompts in txt2img/img2img
generate the Scene, background character/objects & foreground character/objects
generate Text Layers all with toggles
use fixed seeds on each layer to lock in your progress
FAST MODE: toggle dumb/Flux Upscaler for each layer, speeding it up
V10
Flux Inpainting
used Flux with difference diffusion
manual masking in the image loading node
Flux Inpainting SAM2
used Florence2 LLM for prompt 2 mask
used SAM2 for segmentation into FLUX
same as above workflow for inpainting side
V11
Flux Inpainting SAM2 CNET
Flux Inpainting SAM2 from V10 with Lora & CNET Depth
Flux Inpainting SAM2 with Depth+Canny+HED
Flux Inpainting SAM2 from V10 with Lora & CNET depth+canny+hed stack
Flux Lora CNET
Flux txt2img/img2img with Lora & CNET depth+cenny+hed stack
Flux Enhance
Flux Enhance
img2img enhancer, using Florence2 detailed caption conditioning
V12
Flux Scene Editor
Scene Creator with image loading layers instead of generation.
faster, but requires images be made in advance (asset library)
Flux Video Detailer
intended to fix bad AI video details or blurring
denoises with boost, no upscale, tiled 320px
video to video
Flux Live Portrait
Loki-LivePortrait method, with Flux bridge feeding the input
animates a generated image, with a talking video driver.
Flux Mimic Motion
Loki-MimicMotion method, with flux bridge feeding the input
animates a generated image, using a video input.
V14
Flux Lora Tester
Creates two images with the exact same parameters for comparison
Lora ON vs Lora OFF
outputs all images save to sub folders by lora name
outputs specially formatted comparison image, (side by side split ON/OFF)
Flux Lora Compare
Load up to 8 Intermediate Lora models
Creates images with the exact same parameters for comparison
Lora Off output at the top for baseline
each stage can be toggled easily
outputs specially formatted comparison image, with Epoch labels
see demo images for this pack version
V15
Flux IP Adapter Basic
uses the new Flux IPAdapter
Flux IP Adapter Lora
Adds Lora Stack to the workflow
Flux IP Adapter Lora CNET
Adds CNET to the workflow
Compare versions
all the above 3 IPA workflows
4 outputs (foda standard, IPA off & two IPA stengths)
These 4 images create a presentation Grid for comparison
V16
Quality of Life Update (28 Workfows in Pack)
re-organised Workflows by Function group
introduced "Easy Reroute" system for Model/Clip/VAE
allows easy connections to alternate Flux models (FP4/GGUF)
all previous workflows stored in /_archived/
V17
Foda Flux Nova Relighting
generated image with flux CNET/Lora/IPA into Nova IC-Light
includes Foda Flux Animation Enhancer, special for detailing/upscale
Foda Image Nova Relighting
same as above, but faster, due to skipping generation, loading images instead
includes Foda Flux Animation Enhancer, special for detailing/upscale
All packs include the previous versions, new workflows will be added as more capabilities are unlocked. When we reach V20, all workflows will be "refactored" to clean the pack up.
You will need:
https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true
place both inside ComfyUI/models/clip/
VAE
ae.sft (old model, still works, now below as safetensors)
or
https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true
place inside ComfyUI/models/vae/
Diffusion Model
https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.sft
place inside ComfyUI/models/unet/
New Comfy Checkpoint here:
Schnell GGUF Models here:
https://huggingface.co/city96/FLUX.1-schnell-gguf/tree/main
(GGUF T5 text encoder here:)
https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf/tree/main
many more configurations are possible, use the reroute system to connect whatever you wish to use !
CONTROLNET
https://huggingface.co/XLabs-AI/flux-controlnet-collections/tree/main
Flux Canny & Depth V2 were released !
Clip Vision L
https://huggingface.co/openai/clip-vit-large-patch14/tree/main
place inside ComfyUI/models/clip_vision/
consider renaming to clip_vision_l.safetensors
IP ADAPTER
https://huggingface.co/XLabs-AI/flux-ip-adapter/tree/main
flux-ip-adapter.safetensors
place inside ComfyUI/models/xlabs/ipadapters/
because IP Adapter is so new, the nodes may not be available yet:
https://github.com/XLabs-AI/x-flux-comfyui
but install just fine :)
SAM2 Segment Anything
https://github.com/kijai/ComfyUI-segment-anything-2
Florence2 LLM (prompt 2 mask)
https://github.com/kijai/ComfyUI-Florence2
LivePortrait or MimicMotion, setup instructions are here:
https://civarchive.com/models/539936
Kijai's FLUX1 Comfy Lora's
https://huggingface.co/Kijai/flux-loras-comfyui/tree/main/xlabs
Faetastic Details Lora
https://civarchive.com/models/643886
Need LUT's in cube format for color correction?
Description
FAQ
Comments (7)
***FIXED - see further below:
Getting the same message for this and multiple other workflows...
"Warning: Missing Node Types
When loading the graph, the following node types were not found:
CheckpointLoaderNF4
No selected item
Nodes that have failed to load will show as red on the graph."
Reinstalling the "bitsandbytes" node makes no difference.
Reinstalling the associated GGUF node makes no difference. Honestly... I've just about given up on trying to get this to work and annoyingly, it's in almost all recent Flux workflows.
** for anyone else getting this... my problem was doing this manually...
As per the note on the workflow.. switch the channel to dev --> "install missing custom nodes" --> do the "try fix" for this node --> restart --> should be working now --> switch channel back to default.
hi there, I do not use the Checkpoint loader NF4 in any workflow, because i never downloaded it :)
I have no idea why you are seeing this, perhaps that loader was saved to one of the workflows accidentally.
Certainly it's not being used, you can remove it
I only use the UNET loader by default, although everyone is free to use which version of Flux they so choose.
I explained this a bit in the V15 video, it's something i have to say over and over each video, because there are so many versions of quantized flux now. I have been using the same launch schnell/dev for the whole time.
have fun !
What is model shift. I can see how you are playing with different things and how it's affecting outputs. But now I'm more confused where to use controlnet/ipadapter/modelshift/lora? what they are meant to do? any wiki I can look into?
every pack has an associate tutorial video, if you watch them in order you will learn everything you need to know. each previous pack is contained in each later version, so you only need the latest download.
The Video number corresponds to the pack version.
The workflow pack series is also a free e-learning course covering everything you need to know involving Flux with ComfyUI.
TLDR - Model Shift essentially boosts the denoising, effectively giving more control over the outputs and access to extremely low denoising, so the image stays more similar while still being visibly transformed by the prompt.
@driftjohnson Model shift is more common in live model merging and we decide strength to give to each model. Here, since we are only using one model i.e. flux it's probably refers just model strength
The rate at which you output cool workflows and videos is high! Much respect!
Are you a machine?
Yes.