I made a small free app where you can use my lora with the flux. Click if you wanna check it.
(Updated both versions, both 0.1 and ComfyConverted are the same file now. please redownload.)
Deep Semantic Comprehension: Flux+ Detailer excels in understanding and representing the full spectrum of your paragraphs, capturing every detail and emotion.
Curated Dataset: Powered by an exclusive, handpicked dataset, Flux+ Detailer delivers quality and precision with every generated image.
Our curated, high-quality dataset powers Flux+ LoRA, enabling it to improve the paragraphs by both adding depth and detail.
V0.2 Comparisons: https://imgsli.com/Mjg4MDIy
https://imgsli.com/Mjg4MDIw
https://imgsli.com/Mjg4MDIx
V0.1 Comparisons:
https://imgsli.com/Mjg2MjU4
https://imgsli.com/Mjg2MjU5
I organize and clean my captions with https://texttidy.app
Nux's workflow for Flux+ LoRA: https://gist.github.com/Beamhi/28c3d44fcc479a82f06cc0e43a784fec
Another Workflow: https://civarchive.com/models/618997/simpleadvanced-flux1-comfyui-workflows
Checkpoint: https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors?download=true
Trained by ogkai
Small Note: Sending my regards to that Allen guy from upwrk.
Dearly thanks to https://civarchive.com/user/nux1111 for testing the model and generating amazing images.
Description
512px 500 high quality img text pairs,
Will update with 768px version soon.
And fine tune is on the way.
FAQ
Comments (22)
i have lots of error when loaded, type:
lora key not loaded: double_blocks.9.processor.qkv_lora2.up.weight
etc....
and the lora has no effect to this image
i have the same errors.Tryed it with all three dtypes and the standard checkpoint,
You might need to convert it... https://huggingface.co/comfyanonymous/flux_RealismLora_converted_comfyui/commit/cbe0127f8e234b01320cf9094b06b990ac62a1a7
@ericseipel438 this is converted. I am checking if I uploaded the wrong version.
Update boys. You were right. I confused the files.
No, lora simply doesn't work; it shows lots of weight errors when loading, and the images with and without lora look exactly the same
@Yume_ Please redownload the lora and update comfyui. It does work.
No errors for me anymore. I have to figure out how it works the best now.Which strength could it be?
@ogkai_1111 thank you!
@Redtiger85 My good friend Nux is uploading his workflow on Civitai when he tested my lora momentarily. I will be sharing with you guys. You can try on that as well. For the best parameters, feel free to help me out. We are starting a fine tuning on this and I don't have much time to do tests! Thank you!
Excellent work! I'm getting good results at 0.5, tends to clean up small details and bring some more coherence to the image.
checking boys.
Update boys. You were right. I confused the files.
workflow for the lora https://gist.github.com/Beamhi/28c3d44fcc479a82f06cc0e43a784fec
dead
The lora is useful to clean up images. Thanks, It works better than the realistic lora.
Works fine, keep them coming.
bad result in comfy with person lora
v0.2 is coming tomorrow.
GOAT lora. excellent quality
New version is out.