I have trained Flux Dev on my SDXL dataset and merged loras, correcting anatomy censorship and excessive bokeh/blurred backgrounds.
V12 - Merge with SRPO model and some loras, a bit cleaner looking than previous models, might want to use some grain loras if you want that amateur look.
I think not as good at NSFW so might want to use my lora for that: https://civarchive.com/models/652791/jibs-flux-nipple-fix
Jib Mix Flux v8-Flash! SVDQuant-4bit
Update 06/04/2025 - NOW WITH CONTROLNET AND NATIVE LORA SUPPORT!
This new SVDQuant format currently requires ComfyUI and an Nvidia 2000 series+ GPU.
I recommend;
Guidance scale = 2.5-3.5
Sampler = dpmpp_2m.
Scheduler = Sgm_uniform, Beta or Custom Stigmas.
This model is a new format that is a little tricky to first get setup but it is worth it as you can make flux images super fast (5 seconds on a 3090, 2.5 sevonds on a 4090 and 0.8 seconds on 5090! at 10 steps) ,
You need to install the nunchaku project following the instructions there.
and the nunchaku ComfyUI custom nodes to get it to work.
Download and unzip the archive from Civitai to: \comfyui.git\app\models\diffusion_models\jib-mix-svdq\
This is the Nunchaku workflow I am using: https://civarchive.com/models/617562
Thanks a lot to theunlikely for running the Quantization which takes an H100 GPU 6 hours!
Unfortunately, it seems the NSFW capabilities are reduced by the quantisation but can be bought back with my NSFW Lora you need to use the Nunchaku lora loader in the node pack and it will auto convert the lora for use the first time (and optionally can be saved).
There is currently a size limit above 2 million pixels 1024 x 2048 that causes a crash in ComfyUI, so if you get that lower your gen/upscale resolution, The Developers say a fix will be released this week.
Jib Mix Flux Version 8 AccentuEight
Much better Skin texture than Jib Mix Flux V7 but without the bad Flux Lines of Jib Mix V6.
NSFW anatomy is slightly lacking so I have uploaded a separate NSFW model that has very slightly reduced details / artstyles or loras can be used.
The V8 Pruned Model nf4 (6.33 GB) model is actually the Q4_0.gguf
The V8 Pruned Model bf16 (11.84 GB) model is actually the Q8_0.gguf
15/04/2025 - The new fp16 version of V8 provides very slightly better details and anatomical and image consistency, if you have the VRAM for it....
Jib Mix Flux Version 8 AccentuEight NSFW
Better female anatomy with a slight loss in skin details.
The V8 NSFW Pruned Model nf4 (6.33 GB) model is actually a Q4_0.gguf
Jib Mix Flux Version 7.8 Clear Text Focus
This version focuses on having more readable text generation.
The Skin texture may not be generally as realistic as Jib Mix v6 or v7.
It has less nipple flashing through clothes.
The 2000s Analog Core @0.6 weight, removes some of the plastic Fluxness of this merge without hurting the text.
Jib Mix Flux Version 7.2 Pixel Heaven
7/7.2 mainly fixes "Flux Line" caused by merging some effected lora's, this has caused some drop in photo realism but an increase in drawn/concept ability and general details.
It also tones down obsessive amounts of freckles (especially red freckles).
The V7.2 Pruned Model nf4 (6.33 GB) model is actually the Q4_0.gguf
The V7.2 Pruned Model bf16 (11.84 GB) model is actually the Q8_0.gguf
Jib Mix Flux version 7 PixelHeaven - beta
The main change is it removes "Flux lines" that plagued V6 and the original Flux Dev to some extent.
It may be overdoing freckles a lot but I wanted to see what people think of it. hence the Beta name.
I really recommend using Movie Portrait Lora on quite a high weight for a less plastic look, but I couldn't merge it into the model as in testing it was a Lora that can cause "Flux Lines"
Jib Mix Flux Version 6.1 Real Pix Fixed
6.1 mainly tries to fix small stubby hands or massively distorted hands/arms.
(If you still have problems with hands and you are using a low step count around 8 then increasing the step count usually fixes it alternatively applying a low weight (< 0.10) of the Hyper Flux lora pretty much always fixes hands, although you may find it lowers details).
V6.1 is more realistic on the CFG / more natural faces.
I think it actually does art/cartoon styles a bit better than the original v6 as well.
The V6.1 Pruned Model nf4 (6.33 GB) model is actually the Q4_0.gguf
The V6.1 Pruned Model bf16 (11.84 GB) model is actually the Q8_0.gguf
V6 still has the most detailed backgrounds in my testing.
The V6 Pruned Model nf4 (11.84 GB) model is actually Q8_0.gguf
The V6 Pruned Model bf16 (6.33 GB) model is actually a Q4_0.gguf
Jib Mix Flux Version 5 - It's Alive:
Improved photorealism. (Less likely to default to painting styles)
Fixed issues with wonky text.
More detailed backgrounds
Reconfigured NSWF slightly
fp8 V4 Canvas Galore:
better fine details and much better artistic styles, and improved NSFW capabilities.
fp8 V3.0 V3.1 - Clarity Key
I initially uploaded the wrong model file on the 21/10/2024, it was very similar but the new file since 22/10/2024 has slightly better contrast and was used for the sample images.
This version Improves detail levels and has a more cinematic feel like the original flux dev.
reduced the "Flux Chin"
Settings - I use a Flux Guidance of 2.9
Sampler = dpmpp_2m.
Scheduler = Beta or Custom Stigmas.
FP8 V2 - Electric Boogaloo: Better NSFW and skin/image quality.
Settings:
I find the best settings are a guidance and 2.5 and a CFG of 2.8 (although CFG does slow down the generation).
When using Loras these values may/will change.
Version: mx5 GGUF 7GB v1
This is a quantized version of my Flux model to run on lower-end graphics cards.
Thanks to @https://civarchive.com/user/chrisgoringe243 for quantizing this, it is really good quality for such a small model.
There are larger-sized GGUF versions available here: https://huggingface.co/ChrisGoringe/MixedQuantFlux/tree/main
for mid-range graphics cards.
Version 2 - fp16:
For those with high Vram Cards who want maximum quality I have created this merge with the full fp16 Flux model. if you "only have 24GB of Vram" you will need to force the T5 text encoder onto the CPU/System RAM with this force node on this pack:
https://github.com/city96/ComfyUI_ExtraModels
Those people waiting for a smaller quantized model I am still looking into it.
Version 2:
Merged in 8 Step Hyper Lora and some others.
Settings:
I like a Guidance of 2 and 8-14 steps.
Resolution: I like a around 1280x1344
Version 1 : brings some of the benefits and look of SDXL with the massive prompt adherence benefits of Flux.
Settings:
I like a Guidance of 2 and 20-40 steps.
Description
Mainly fixed having too many skin freckles in V7 Beta and pulls back more towards photo realistic.
FAQ
Comments (35)
Does anyone know why the model (Jib Mix Flux V7) won't launch on Forge?
It works, you need to have VAE/CLIP/T5.
@REgenerationSD Possible to elaborate on that? Please
@Kickingbull222 Here is a guide to setup the separate parts in Forge: https://civitai.com/articles/7828/simple-setting-forge-ui-for-runing-flux-no-include-vae-and-clip
So if this model really does remove "Flux lines", then it can be used as an upscaling model, because the Flux lines appear only when going over a certain resolution
It does massively reduce them from my previous models for normal sized images https://civitai.com/images/50115978 but they are still there slightly when you produce very large sizes natively, here I did a nearly 3,000px text to image https://www.reddit.com/r/FluxAI/s/t8qjSXpQ9F
The Flux lines are quite subtle but still there.
cool
Any chance of a Q8 for the latest versions? I liked V6 outputs, but the flux lines kept me from continued use. Thanks!
Sure, I have been meaning to do that, should be uploaded in around 1 hour.
Sorry, I got distracted, it will hopefully be created and uploaded tomorrow, in about 16 hour from now.
Awesome. No rush, and thanks!
@Unseelie_AI The Q8_0 is now here https://civitai.com/api/download/models/1249737?type=Model&format=GGUF&size=pruned&fp=bf16
@J1B Awesome! About to download now! Thanks!
I am having trouble keeping the outcome SFW, as soon as someone is busty, bare breasts are there, even if the prompt says they are wearing a dress. Is that intentional or is there a specific way to avoid NSFW with this otherwise amazing checkpoint?
I know they say you're "not supposed to use a negative prompt" with Flux, but sometimes I just have to. Try making the CFG 1.1 and adding in a negative prompt "nude" or "NSFW". See if that helps. ;) (Btw, this will slow down your processing significantly, but if nothing else works...)
Give less emphasis to your busty breasts by inserting a bra under the dress or by inserting details of the dress.
I found a seemingly working solution. I used https://civitai.com/models/660029/nakednudeforfluxsevenof9nsfw Lora with a negative weight (-0.4 worked well). I will test it a little more
Alright, even with the same prompt as before, negative weight "naked" Lora works like a charm!
@DerpHerp112 Glade you have sorted it. You could also try my nipple fix Lora at Negative value:https://civitai.com/models/652791/jibs-flux-nipple-fix
as my model has some of that mixed in to correct anatomy.
It is a fine balancing act between messed up NSFW anatomy and always having nudes with Flux because of the way they censored the base models I think.
Do you have any recommendations for using ControlNet with your checkpoint? Settings I use with the base Flux model don't seem to work with this one.
hmm, I have not tried it to be honest, I will have a test later. I have been meaning to post a new version of my workflow with WaveSpeed integrated anyway.
know it's odd but any recommendation on running your gguf q8 dev model with 4 steps ? somwthing general about flux i know but need to hear from you an advice .
You will not be able to get reliably good images at 4 steps, maybe 1 in 4 images might come out alright depending on the prompt, if you go up to 6 steps it will look better more reliability, but personally I don't go below 18 steps as I like the best quality.
you can always try it with the 8 or 16 step loras, quality does drop a wee bit but i'm sure it will give better results than trying to run a 20 step model at 4 steps
Great model! What do you think of the de-distilled version?
I have looked into it briefly, but I am still not really sure how it is done. I already use the Dynamic Thresholding node in ComfyUI if I want to push to a higher CFG values without negatively effecting quality "image burn". It seems quite interesting to dive into the technicalities of it, so I might give it a go.
@J1B This is great news! I am absolutely sure that you will get a great model. I will be waiting impatiently
great brother, where do I have to install this one? comfyui/Storage/models/checkpoints?
The model needs to placed in the folder: /comfyui/models/unet or comfyui/Storage/models/unet - depending on your configuration.
what clip and vae doo i need to use ? i don't really get the diff between the unet folder and checkpoint folder except that i understood that for unet folder checkpoints, you need to use an external clip and vae right ?
@TheProphet22 Yes - you need to use the t5xxl and the clip_l clip models and the Flux VAE.
You can download the workflow, which works fine for me here - https://civitai.com/images/54751606
@TheProphet22 you need the T5 Text encoder (can be the same one used for SD3): https://civitai.green/models/704402/flux-textencoder-t5-xxl-fp8-e4m3fn
And the Clip-L.
If you use my workflow: https://civitai.com/models/617562/comfyui-workflow-flux-to-jib-mix-refiner-with-negative-prompts
You can optionally also use Clip-G in a triple clip loader and it changes the image slightly.
@J1B Thank you for your help ! i use the dual clip with t5xxl-fp8-e4m3fn and clip-L , where to find clip-G ?
@TheProphet22 I have uploaded it here: https://civitai.com/models/617562?modelVersionId=1106432
Love it! <3
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.










