Using (multiple) LoRA in ComfyUI: A Minimalist Workflow
HTML-код
- Опубликовано: 15 сен 2024
- Discover how to streamline your ComfyUI workflow using LoRA with our easy-to-follow guide. Boost efficiency and simplify your projects today!
🖹 Article Tutorial:
- myaiforce.com/...
🔗 Links:
- Running Stable Diffusion on RunPod: runpod.io?ref=jden1933
- ComfyUI LoRA examples: comfyanonymous...
- ComfyUI manager Github page: github.com/ltd...
- Efficiency Nodes Github page: github.com/jag...
This is so wonderful, thank you! This should be the default experience with comfyUI... this is so much more easy to work with. I'm now making exactly what I want without having to drag around and follow a maze of wires.
One of the best ComfyUI tutorials. Amazing workflow.
Wow, thanks!
How I can still previewing the image, but only save when I need it, I don't want to save all the images, only the ones it stands out, is there a why to easy select the image and just save the one that is the best in a batch of 4 generations or more?
Can you build a workflow that incorporates Video to Animation unsampling (just like the one from Inner Reflections) so that the video doesnt morph, but then include a load loras box for style and/or some kind of style reference image because right now I've loaded a loras box into that aforementioned workfllow from Inner Reflections but no matter how hard i try with prompting or certain checkpoints i cannot get it to match the style of my loaded lora. I want my video to be animated and look like a certain Lora.
life saving video, Thank U
Glad it helped!
Great! But how do you use the SDXL loader? It doesn't have any positive/negative conditioning outs.
I'm using a similar workflow as this, and its great having far less nodes on the screen. One issue I'm having is that the lora metadata is not going into my save image with metadata node (everything else is, just the lora info isn't). Is there a way to make the metadata show up on a saved image using a lora stacker? It will help on civitai to show what resources were used in creation so you can support lora creators...
thank you amazing work but what if the lora in the lora stacker needs a different clip skip or no clip skip at all how can i do that ?
THX, can you make a workflow for SDXL? The loader ist dirfrent form the 1.5 loader.
The process for SDXL is similar, only the nodes corresponding to SDXL need to be selected, i.e. the Efficient Loader and KSampler corresponding to SDXL.
does this work for flux?
Yes, this works for Flux because the FP8 version of Flux functions similarly to other checkpoints. You can use the default ComfyUI workflow to integrate it seamlessly.
@@my-ai-force And with GGUF?