COMFYUI SUPPORTING !!!!!!!!!!!!!!!!!!!!!!!

[vuongminh1907]

Hi Team, I'm so thankful that you guys created such a great repo. So we implemented InfiniteYou on ComfyUI 🚀.

Here is our repo link: https://github.com/ZenAI-Comfy/ComfyUI_InfiniteYou

Can you attach this link to share with our community? It's a pleasure to develop with InfiniteYou 🔥

[MegaCocos]

Trying your custom node and get the error on loading ComfyUI "ModuleNotFoundError: No module named ComfyUI_InfiniteYou.ZenID"

[GraftingRayman]

how much ram or vram do we need for this?

[EndlessSora]

Thanks a lot for your contribution! Could you please help add LICENSE and citation BibTeX to your code and model repos? Our code is released under the Apache 2.0 License, and our model is released under the Creative Commons Attribution-NonCommercial 4.0 International Public License for academic research purposes only.

We are also planning an official ComfyUI. Will also link great unofficial implementations. Thanks you so much.

[vuongminh1907]

thanks @EndlessSora, I just added LICENSE and Ciatation. Happy to add our Node to your repos.

[vuongminh1907]

hi @GraftingRayman, @gnusupport tried my node with 3090 with 24VRAM and he was successful.
ZenAI-Vietnam/ComfyUI_InfiniteYou#5

[MegaCocos]

hi @GraftingRayman, @gnusupport tried my node with 3090 with 24VRAM and he was successful. ZenAI-Vietnam/ComfyUI_InfiniteYou#5

@vuongminh1907 I tried your node on 4080 with 16VRAM and it works well, no problem 👌

[vuongminh1907]

That's great @MegaCocos! Can you share the workflow or create a PR in my repo to share with everyone?

[vuongminh1907]

Maybe you use Flux quantization or something else?

[MegaCocos]

That's great @MegaCocos! Can you share the workflow or create a PR in my repo to share with everyone?

I used your workflow, no problem, just changed your load model and TE nodes for GGUF versions. No problem to share, but i am at work now. Still need to add to your workflow the LoRA node and the save image node to make it more usable, but your workflow works as it is. I use FLUX Dev Q8 GGUF model, as usual.

[EndlessSora]

thanks @EndlessSora, I just added LICENSE and Ciatation. Happy to add our Node to your repos.

Thanks a lot! Will do that soon.

[EndlessSora]

@vuongminh1907 Could you please add the Creative Commons Attribution-NonCommercial 4.0 International Public License to your model repo? Thank you so much!

[vuongminh1907]

Sure @EndlessSora, it's added.

[gnusupport]

But why that non-commercial license? That is not a free software license and not open source license. That means also that this custom node is not going to get enough exposure and it cannot be included in many other software. Please understand this.

[EndlessSora]

But why that non-commercial license? That is not a free software license and not open source license. That means also that this custom node is not going to get enough exposure and it cannot be included in many other software. Please understand this.

Hi, thanks for your interest! Sure, we understand this. Our model is trained with FLUX.1-dev. Any model trained FLUX.1-dev must follow FLUX.1 [dev] Non-Commercial License and even more strict ones. CC-NC-BY 4.0 is still open-source but non-commercial. Thanks.

[Hany1978]

@MegaCocos PLEASE share your workflow with the GGUF changes, because I m a beginer in ComfyUI and can not edit and change the current workflow from fp8 to GGUF, so if possible please share it. Thanks In Advance!

[MegaCocos]

@MegaCocos PLEASE share your workflow with the GGUF changes, because I m a beginer in ComfyUI and can not edit and change the current workflow from fp8 to GGUF, so if possible please share it. Thanks In Advance!

@Hany1978 I just uploaded the workflow to CivitAI https://civitai.com/models/1424364 you can download it and use it accordingly the License of Bytedance and ZenAI. The node is made by ZenAI, i only replaced some nodes and added some others, but all of those are minor updates. You must follow all download and install instructions of ZenAI https://github.com/ZenAI-Vietnam/ComfyUI_InfiniteYou and install all requirements accordingly Bytedance https://github.com/bytedance/InfiniteYou

The node was tested on 16 Gb VRAM RTX 4080 Super with the normal speed of 1 - 2 minutes for 1024 x 1024 image generation.

[MegaCocos]

That's great @MegaCocos! Can you share the workflow or create a PR in my repo to share with everyone?

@vuongminh1907 Hello, finally i have uploaded your workflow edited for using GGUF versions of model and TE loader + Lora + Save Image and some minor updates on CIVITAI. And I will add fast selectors for models when I have a bit more time. Hope you don't mind, all the work is yours, my just minor updates. If you need me to change something, just ask. Also you can use my version as you wish https://civitai.com/models/1424364

The generation speed on 16 Gb RTX 4080 SUPER is of 1 - 2 minutes per 1024 x 1024 image.

[MegaCocos]

@EndlessSora Hello, i just uploaded the workflow of @vuongminh1907 slightly modified by me and based on your fantastic model on CivitAI https://civitai.com/models/1424364 as i was asked here by @Hany1978 . I put links to your GitHUB and mentioned License agreement. Now i am on hurry and sleepy but if you need any changes by me on the CIVITAI page, just ask and i will arrange it when i am recovered.

The generation speed on 16 Gb RTX 4080 SUPER is of 1 - 2 minutes for 1024 x 1024 image, completely normal speed by my opinion

[Hany1978]

@MegaCocos Thanks In Advance! for your effort!!!

[MegaCocos]

Added the Switch between SIMILARITY and FLEXIBILITY models (the red box with 0:Similarity / 1:Flexibility options) and made minor adjustments (V 4.0) https://civitai.com/models/1424364?modelVersionId=1613531.

SIMILARITY:

FLEXIBILITY:

[niftyflora]

i was excited to test this and now that its working i just found out its muuuuuuuuch worse than good ol ReActor face swap also 2 minutes slower

[MegaCocos]

i was excited to test this and now that its working i just found out its muuuuuuuuch worse than good ol ReActor face swap also 2 minutes slower

@niftyflora Yea, i agree, for face swapping it is not the best tool, i don´t use it for face swapping. It is for more creative works.