Brie Wensleydale🧀🐭

AI R&D, animation, gals and news.

I make custom workflows and deploy comfy / webui.

Feel free to DM me!
Business inquires go here: [email protected]

[Brie Wensleydale🧀🐭 @SlipperyGem]

New Brie Advanced Character Fusion is now up on both github and CivitAI. Thanks once again to

@tori29umai, @dx8152 as well as the @Alibaba_Qwen team for 2511.

This one is beeeg, and will serve as a base to other workflows.

https://t.co/N5mSv2jeJw

[Brie Wensleydale🧀🐭 @SlipperyGem]

thanks to @arman_beat777 for mentioning it ~

[Brie Wensleydale🧀🐭 @SlipperyGem]

Didn't know about FreeFuse before.

Support SDXL, Z-Image, Flux Klein, and, as of 2 days ago, Krea 2.

As per their description: "FreeFuse enables seamless multi-subject comp by fusing multiple LoRAs without additional training or user-defined masks."
https://t.co/OxFcseuy6y

[Brie Wensleydale🧀🐭 @SlipperyGem]

@arman_beat777 Oh that's super neat! I did not know about FreeFuse! They support a bunch of models too! Including Krea!

[Brie Wensleydale🧀🐭 @SlipperyGem]

Now this is the type of ID preservation type functionality that I crave.

I'm gonna be traveling tomorrow, so I won't be able to test it. I'm curious to see how well it works on stylized characters...
https://t.co/Rij9GpPam2

https://t.co/0aEhK3kOqx

[Brie Wensleydale🧀🐭 @SlipperyGem]

Got a pair of outfit extractor and clothing try-on loras from kingroka working on a set of workflows from nsfwVariant.

It works, I like it, but no time to test more for now.
https://t.co/5S2nMjqeo6
https://t.co/MF4jKsKAlo
https://t.co/A8mNPOPHKz

[Brie Wensleydale🧀🐭 @SlipperyGem]

That ain't mini...

[Jaid @Jaidcel]

MaxiMax

[Brie Wensleydale🧀🐭 @SlipperyGem]

So this is why I'm starting to see Krea 2 loras with editing capabilities pop up...

This will certainly yield more incredible things soon enough...

[Ostris @ostrisai]

Just pushed support in AI Toolkit to train @krea_ai Krea2 with reference images. In my testing, it learns edit concepts just as fast as an actual edit model would. More in 🧵

[Brie Wensleydale🧀🐭 @SlipperyGem]

@nonRealBrandon It is astounding that ID preservation is still so difficult in this day and age with open source models.

Some things like head swap have gotten much easier, but what this does is quite difficult, no?

[Brie Wensleydale🧀🐭 @SlipperyGem]

@arman_beat777 Please @ me if you have cool results from this... thank you

[Brie Wensleydale🧀🐭 @SlipperyGem]

@LovelaceHopeful This ain't my output, but I am trying out various methods to do this type of stuff.

[Brie Wensleydale🧀🐭 @SlipperyGem]

JT88 has made another useful node. Last time was the multi-cam angle node. This is in the same vein but with lights.

I assume this plugs into that nice Flux Klein 9B light control lord I posted a few days back.

[jtydhr88 @jtydhr88]

keep improving, it also can be multi lights support

[Brie Wensleydale🧀🐭 @SlipperyGem]

IdeoKrea lora for Krea 2 that makes it look and also function more like Ideagram 4, including the json structuring apparently.

https://t.co/xSx9Fppi2K

Left: Before
Right: After

Somebody posted this on a thread of mine, but I lost it.

I always lose the things I don't post...

[Brie Wensleydale🧀🐭 @SlipperyGem]

I think I posted this a few days ago, but this pose one was mixed in with the others.

I think this is it. With this, I can finally toss Qwen Edit 2511 aside with AnyPose and other re-posing methods...

I guess only Pixel Smile remains for QIE2511.
https://t.co/SXyeyKcETO

[Brie Wensleydale🧀🐭 @SlipperyGem]

Looks like SCAIL 2 is quite capable when it comes to this type of manipulation. nice ~

[toyxyz @toyxyz3]

Wan 2.1 Scail-2 stress test

[Brie Wensleydale🧀🐭 @SlipperyGem]

@888surf Its not supported on this custom node, but apparently @realrebelai has gotten MrFlow to work, to a certain degree with Krea 2.

[Brie Wensleydale🧀🐭 @SlipperyGem]

MrFlow ComfyUI Plugin allows for serious speed increases.

10x speedup on Qwen, 8x on FLUX.1, 20x on Z-Image
low quality loss

Nothing about Qwen Edit 2511, which is the core of my interests. Hopefully it works for that.

[Wildminder @wildmindai]

MrFlow ComfyUI Plugin
helper nodes and workflows for QwenImage

https://t.co/OEdqZmlWXM

[Brie Wensleydale🧀🐭 @SlipperyGem]

@SteveWarnerFL For client and personal work, when quality is important, we will sometimes lean Qwen, when speed is important, we will always lean Flux. Also, there's still some loras where Flux has no equivalent, like AnyPose.

I'm currently digging into this.
https://t.co/kZahZqVbc2

[Brie Wensleydale🧀🐭 @SlipperyGem]

The standalone version.

While ComfyUI is convenient, for audio stuff that isn't video to audio, I find it unnecessary to run things within comfy, although, it often is the only viable choice.

https://t.co/ixlaWuqGz6

[Brie Wensleydale🧀🐭 @SlipperyGem]

Higgs Audio v3 for ComfyUI. Just got curious and was poking around to see what was the SotA local TTS and voice cloner out there. Seems like this is a candidate.

Saganaki22 also made an even better windows standalone app for Higgs Audio, linked below.
https://t.co/G0cdkB0GCK

[Brie Wensleydale🧀🐭 @SlipperyGem]

Krea2 Depth Controlnet is out.

I wonder if this capability can be combined with the IdeoKrea lora that I posted a day or two ago. Would be neat to get both complex json control with control nets.

[toyxyz @toyxyz3]

https://t.co/obKUYoKNni