A pretty simple workflow with a few common features. There are some important (more or less) notes inside the workflow. It is designed for 'legacy' mode. For now, I still think that Nodes 2.0 are mostly buggy and slower garbage, even though they have a few cool things. The metadata of a final image will not contain the diffusion model name; I still haven't found a good way to add it to the metadata. Fixed in v.1.2
I didn't notice that Lora doesn't get into the metadata for automatic detection. Fixed in v2.0.
Description
v3.5:
Remaded bookmarks, added a disclaimer about bookmarks.
Added save image node without meta; this node wipes any meta data including workflow.
Added toggle for both types of saving images. You can choose to disable them both.
"Fixed" previews inside subgraphs. "Discounted" is not very elegant but works.
Few small visual and internal changes.
Requared my extension v2.5.0 or newer.
v3.6:
I replaced the boolean toggle node with my own. An easy-to-use extension is no longer needed for this workflow.
FAQ
Comments (5)
Thanks for this amazing workflow, the txt2img works great for me so far! Just have one problem, when I enable the ControlNet option and add an image of a character t-posing, the output I get is just a gray box. When I disable ControlNet, it works again. Any idea what I do wrong? For info, I set the "preprocessor" to "DWPreprocessor".
Thank you for your kind words. I'm not sure if I understand what you mean by "to get grey box," so I apologize.
If it happens at the step of preprocessing, that shouldn't be the case. If it occurs at the end of generation (I mean the final image), that's a different issue, like the main reason why Z-image controlnet doesn't support it. I think nobody besides SD1.5 really works with it (probably only Klein, but not directly as controlnet).
The supported preprocessors are Canny, Depth, Pose, MLSD, Hed, and Scribble. The reason I use the node for preprocessing is that it's universal for many other models, but that doesn't mean each model supports all of those preprocessors.
I tried using different methods myself, and it either had no effect or broke the final image. So try to use Canny or Depth — that's all I can suggest. If I still don't understand you correctly, please let me know.
@DemonAlone It happens as soon as the Ksampler starts working, then you can see the box in the preview as you can see on this screenshot: https://i.imgur.com/7Dy5RTR.png
Every output becomes this solid box as long as ControlNet is enabled. I am using the "z-image-turbo-fun-controlnet-union-2.1.safetensors" model for Controlnet.
@Eshinio I see, I didn't try that version. Now I have, same behaviour. Well, the official one (link in my note https://huggingface.co/alibaba-pai/Z-Image-Turbo-Fun-Controlnet-Union) still works and is three times less in size. Have you ever tried it?
@DemonAlone Ah, using the model you linked to seems to fix it! I get a bit weird/ugly results, but I think it's because the ControlNet resolution is different from my generated image resolution. Thanks for the help! :)