ComfyUI - Flux Inpainting Technique
1.0
1 reviews
Description
FLUX is an advanced image generation model, available in three variants:
- FLUX.1 [pro] for top-tier performance,
- FLUX.1 [dev] for efficient non-commercial use,
- FLUX.1 [schnell] for fast local development
These models excel in prompt adherence, visual quality, and output diversity.
Click on below link for video tutorials:
Node Diagram
Discussion
I would like to ask where to download these two models t5xxl fp8 e4m3fn.safetensors,fux1-schnell.sft
Here are the links for those two models. Hopefully this is what you were looking for.
https://huggingface.co/comfyanonymous/flux_text_encoders/tree/main
https://huggingface.co/black-forest-labs/FLUX.1-schnell/tree/main
HI I am getting an error in Comfy, screenshot here: https://i.imgur.com/sOMk9TO.png
It seem it points to Load Vae node, it gets red in the ComfyUI. Is ae.stf from flux-dev and flud-schnell the same? I've downloaded the one from flux-dev, but in this workflow flux-schnell is used. Any tips how to solve this are welcome.
Your other workflow for flux anime is working without any issues for me with flux-dev. What do you think could be wrong here?
20 steps for Schnell??
more steps = higher quality = lower speed. Its personal preference. Play around with it and find what is best for your use case
Why do I get the same result every time?
did you found out why?
set your control_after_generate parameter in the RandomNoise node to random instead of fixed
Thanks for sharing the workflow, I tried to run it but I'm getting this error, any idea what might be wrong?
```
Error occurred when executing LoadAndResizeImage: Invalid background color: image
```
If I set a hex color the workflow works, but the result is not what I'd want, it basically renders the selected color around the area that is inpainting 🤷♂️
use a normal "load image" node instead and then right click and edit in MaskEditor
Hi, thanks for your model. In my "Load and resize image", I have a background_color parameter. This causes me to have an outline on the final image. Do you know how to remove this parameter or give it a neutral value please
anyone getting the color issue in the "Load and resize image" node should just use a normal "load image" node instead and then right click and edit in MaskEditor
hi friend, i used "load image" node, it worked, but the result is black screen
I test your flow and when start i have:
Trying to convert Float8_e4m3fn to the MPS backend but it does not have support for that dtype.
test on mac
Same here. After changing weight_dtype to default, it started to work, but it generates, at best, some strange clothes totally different from the prompt. At worst, it generates something random that doesn’t fit the scene at all.
Nodes missing and can't find them in manager:
- Anything Everywhere
- ImpactGaussianBlurMask
- LoadAndResizeImage
wow, how many vram does it require? Cant run on 16gb gpu
LoadAndResizeImage
Invalid background color: image
Thank you! is perfect! 😍
提示词不起作用，为什么呢
a girl in a street
Node Details
Primitive Nodes (4)
Anything Everywhere (3)
FluxGuidance (1)
Custom Nodes (16)
ComfyUI
- BasicGuider (1)
- SamplerCustomAdvanced (1)
- VAELoader (1)
- UNETLoader (1)
- DualCLIPLoader (1)
- CLIPTextEncode (2)
- InpaintModelConditioning (1)
- DifferentialDiffusion (1)
- VAEDecode (1)
- PreviewImage (1)
- RandomNoise (1)
- KSamplerSelect (1)
- BasicScheduler (1)
- ImpactGaussianBlurMask (1)
- LoadAndResizeImage (1)
Model Details
Checkpoints (0)
LoRAs (0)