[Flux ControlNet] Add support for de-distilled models with CFG

Flux with Inpaiting and ControlNet currently yields bad result with the base model.

To echo this comment, using de-distilled models could potentially help getting better outputs.

Currently the Flux ControlNet pipelines do not support models like https://huggingface.co/nyanko7/flux-dev-de-distill

Anyone wants to contribute with this? This should be a community pipeline in the meantime until we find if the new models are good and get popular, but this could produce better results than the current ones.

Also maybe someone can test this in ComfyUI and see if it makes a big difference.

Anyone wants to contribute with this? This should be a community pipeline in the meantime until we find if the new models are good and get popular, but this could produce better results than the current ones.

Also maybe someone can test this in ComfyUI and see if it makes a big difference.

will take this up 😊

@ighoshsubho

Hey, @ighoshsubho wanna collab on this?

@ighoshsubho

Hey, @ighoshsubho wanna collab on this?

sure

Anyone wants to contribute with this? This should be a community pipeline in the meantime until we find if the new models are good and get popular, but this could produce better results than the current ones.

Also maybe someone can test this in ComfyUI and see if it makes a big difference.
yes i would love to contribute on this

@ighoshsubho

Gentle ping @ighoshsubho

@a-r-r-o-w

@a-r-r-o-w I apologize, due to unexpected time constraints with my primary work commitments, I won't be able to give this the attention it deserves. Please feel free to reassign this to another contributor who can dedicate the proper time and effort to implement this feature.

I will give this issue a look, and I will start with testing this out in ComfyUI

@a-r-r-o-w

Hey @a-r-r-o-w @asomoza , is this one still open for contributions? If yes then I can give it a try!

@a-r-r-o-w

Hiii @a-r-r-o-w @asomoza , Is anyone working on this one or is it still open for contributions ? If yes, then I want to give this a shot !!!

@asomoza

Anyone wants to contribute with this? This should be a community pipeline in the meantime until we find if the new models are good and get popular, but this could produce better results than the current ones.

Also maybe someone can test this in ComfyUI and see if it makes a big difference.
Hi @asomoza,

Wanna give it a try, please let me know if someone is already working on this?

Hey!, i would love to contribute to this issue, can you please assign this to me

Hi! I looked into this on current main. FluxControlNetPipeline gained true_cfg_scale / negative prompt support at some point after this issue was opened, so de-distilled models should work for text2img now. However, FluxControlNetImg2ImgPipeline and FluxControlNetInpaintPipeline still lack it — and the inpainting case is what this issue originally reported.
I've ported the true CFG implementation from FluxControlNetPipeline to both pipelines (matching its conventions, including reusing the positive ControlNet residuals for the uncond branch) and added regression tests. Opening a PR shortly.

@LeonxLJX

I'd like to take this one ([Flux ControlNet] Add support for de-distilled models with C). I'll look into the root cause and follow up with a PR. (claiming via @LeonxLJX)

@LeonxLJX

I'd like to take this one ([Flux ControlNet] Add support for de-distilled models with C). I'll look into the root cause and follow up with a PR. (claiming via @LeonxLJX)

Hello guys, I'm new here. Could I be any help? I would like to contribute in an issue (first issue)

Navigation Menu

Repository navigation

[Flux ControlNet] Add support for de-distilled models with CFG #9635

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[Flux ControlNet] Add support for de-distilled models with CFG #9635

Description

Activity

asomoza commented on Oct 10, 2024

ighoshsubho commented on Oct 10, 2024

charchit7 commented on Oct 10, 2024

ighoshsubho commented on Oct 10, 2024

taneeshamadhu18 commented on Oct 11, 2024

a-r-r-o-w commented on Nov 20, 2024

ighoshsubho commented on Nov 20, 2024

KarthikSundar2002 commented on Dec 2, 2024

M0315G commented on May 11, 2025

shashwatrajiitd commented on Aug 28, 2025

narendramall commented on Mar 22, 2026

harshil-1402 commented on Jun 18, 2026

Gavin-Morris-04 commented on Jul 8, 2026

LeonxLJX commented on Sep 1, 2026

LeonxLJX commented on Sep 1, 2026

mpferreira003 commented on Oct 23, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions