ComfyUI Impact Pack - TwoAdvancedSamplersForMask

Similar to the existing TwoSamplersForMask, you can apply separate KSamplers to the masked area and the area outside the mask. However, TwoSamplersForMask applies the sampling process for the mask area only after the sampling process for the background is complete.

On the other hand, TwoAdvancedSamplersForMask applies alternately at each step, allowing for smooth blending even if the denoise is set to 1.0. With this node, you can apply separate Lora to each area.

[dylanhaze]

Should have a option sampler for backround and use the overlap_factor on that. (to avoid mixing faces when using 2 loras on different characters)

[quietek]

Could you give a link for this workflow? I still don't understand how it work but I think this can solve my problem.

[drltdata]

You can download from ComfyUI from here:

And ComfyUI Impact Pack extension from here:

[tracingzimpossible]

i downloaded this workflow, while "2sampler for mask" is working as shown in the demo but "2 advance sampler for mask "is not working ..after Q prompting all its giving is noise, where im doing wrong ?? any clue??

[Danted]

Hello! Could you help me? I am using the node, TwoSamplersForMask, with a simple box mask covering half the picture. I want to generate two characters, one on the left, one on the right. And it works, but somehow, it looks like two images cut and pasted together ^^'. Like, imagine grabbing an image of a character, an image of another, and putting them next to one another.

Do you know why? How do I get rid for that feeling of "two images put together" and get a single, unified picture?

[patchper]

What is to be filled in the positive prompts in Ksampler if I want to upscale the resulting image/latent? In the case of video, the one with chunli is obvious, but what if 2 different loras is applied in base sampler and mask sampler? Any suggestions?

[yeon_sunn]

Are you using that new module AIT(AITemplate)? Can it increase the speed of generating pictures?
Why did my comfyui report the error directly：'BaseModel' object has no attribute 'latent_format'