FLUX 1 Tools Redux - Tutorial Guide - An IPA-Alike Model From Black Forest Labs Officially

[@kalakala4803]

Thanks Ben! We will integrate this into our work.😊

[@TheFutureThinker]

Let's do it 😎👍

[@Akozehalus]

Thank you for this nice tutorial :) yea inpaint will be nice to weatch also :) thanks man, rly help me out!

[@TheFutureThinker]

No problem!

[@sudabadri7051]

Great breakdown mate

[@TheFutureThinker]

Glad you enjoyed it

[@ovideotube]

so great. could you make a tutorial to use redux to convert a photo to a comic book illustration preserving the subject face identity?

[@haljordan1575]

Can the inpaint tool use a lora to inpaint? Like when you have a character lora and you wanna add it to a scene

[@TheFutureThinker]

Yes , will try with the Flux tools Fill aka inpaint model

[@mikiex]

The depth and canny are also smaller loRAs they seem to work quite well, I needed to lower the loRA strength

[@TheFutureThinker]

Yup :) will try that next

[@TECHTOUR]

can we combine depth and canny? just like sdxl

[@daniellohneiss2789]

It broke my comfyui it says it errors with :ComfyUI_windows_portable>pause. Then it’s just auto closed. That was after updated the python files. I’m thinking it’s too much vram for my 16gig card? An advice to reverse this, or just delete comfy and reinstall 😢

[@JolieKim2795]

hey Ben does it work with other ckpt such bnb nf4, guf. My PC is so slow for fp8 dev 11gb

[@TheFutureThinker]

I try with gguf , it works

[@bobobaba2080]

I followed your instructions perfectly but, I have been getting this error (CLIPVisionLoader
Error(s) in loading state_dict for CLIPVisionModelProjection:
size mismatch for vision_model.embeddings.patch_embedding.weight: copying a param with shape torch.Size([1152, 3, 14, 14]) from checkpoint, the shape in current model is torch.Size([1024, 3, 14, 14]).) all day, I tried to troubleshoot it for hours, but I couldn't fix it.
a little help please

[@TheFutureThinker]

02:49

[@bobobaba2080]

@@TheFutureThinker Thank U so much brother. It worked.

[@flankechen]

generally, I think redux is much too strong img conditioning

[@user-fx9wf9jy4g]

you will be rich :)

[@Kurie3D]

which folder do we stick flux1-depth-dev.safetensor?

[@philippeheritier9364]

Thanks for the excellent tutorial as usual, but I have an error with "sigclip_vision"
CLIPVisionLoader
Error(s) in loading state_dict for CLIPVisionModelProjection
size mismatch for vision_model.embeddings.patch_embedding.weight

[@CarlosDiaz-t2j]

Same error here, I have the same set up with the same clip Vision but didn't work

[@ronbere]

same error

[@ChrissyAiven]

I get a lot of errors too: size mismatch for vision_model.embeddings.patch_embedding.weight: copying a param with shape torch.Size([1152, 3, 14, 14]) from checkpoint, the shape in current model is torch.Size([1024, 3, 14, 14]).

[@piyushjadav533]

Download sigclip from Google

[@Darkwing8707]

@@ChrissyAiven Make sure you update ComfyUI.