I think with the latest ComfyUI update the florence is broken.

Failed to validate prompt for output 72:
* Miaoshouai_Tagger 300:

Absolutely stellar work again, best workflow out there. Interestingly I'm getting way better results with your older workflow using the SDXL refiner. I use juggernaut XL.

I've mirrored settings and not sure if it's my lora though seems unlikely. Regardless it's a fantastic workflow, will need to play around a bit to see if it is the absence of the SDXL model or not causing me the issue.

face_yolov8m-seg_60 in what folder should be place?

This version is a Masterpiece! Thank you for sharing!

Thanks for sharing this workflow. It's been giving me great results.

Theres one thing I can't wrap my head around, how is the img2img denoise and Max Shift, Base Shift interacting with each other?

Any tips regarding these settings would be much appreciated :)

Great workflow.. I didnt 2ndpass Noise injection , Lucy Sharpen and Shadow Highlight v2 in the updated workflow.
Particularly i found 2ndpass noise injection a great feature.

I like the look of this workflow, but I'm having trouble getting it to run. Even with everything turned off, it's still giving me errors about UltimateSDUpscale.

The error says "Input type (float) and bias type (struct c10::Half) should be the same"

Any idea how to fix that?

with the latest comfyUI update, the workflow broke. Something that has to do with the "SAMPLER IMAGE" node. At least for me, i don't know if anybody else has any problems. Would you mind taking a look? Or could you please upload a screenshot of your settings in that node? Because when i load it, the settings are a bit messed up after the update

Great update like always, I noticed and this may be user error that the workflow won't run unless you have an image loaded in the load image node. This is irrespective whether you have img2img set true or false. Any idea what is going on there?

I think the USDU nodes appear to be broken. (#3840 UltimateSDUpscale) is all red and (#3749 UltimateSDUpscale) is outlined red. Sorry if I'm wrong but I'm still learning and so low on energy right now I'll ask for some babysitting lol. The initial gen is awesome though, the workflow is sick. Though it's taking me 60s/it I assume because of my subpar rig(16GB RAM, 1660s(6GB VRAM)). With another SDXL workflow I tried a fully upscaled(4k) image takes me ~10 minutes lol. I've struggled with getting flux workflows to work.

I think this is broken at the moment

If you've installed SAM2 and LayerStyle, they will be in conflict with each other. I confirmed this by running the following command find ~/ComfyUI/custom_nodes -type f -iname "*sam2*" go to your custom_nodes folder and delete ComfyUI-SAM2, and make sure LayerStyle is installed. Then you should be good.

i`m having a issue with Ultimate SD Upscale, it`s installed, but when i open the workflow it`s missing a node, in the Custom Node Misssing tab show the thing, there is a button fix, but nothing happens, anyone can help to fix it?

how to turn off florence??

I'm getting:
SegmDetectorCombined_v2 'NO_SEGM_DETECTOR' object has no attribute 'detect_combined'
but am not sure where I can troubleshoot the issue. The box is outlined in purple

I've installed jiter,still shown '''No module named 'jiter.jiter' ''' .Does anyone know why?

Bro, thanks for the best workflow! Please tell me, some nodes are black, and it is not possible to configure the parameters. I would really like the network control to follow the example more closely when the photo turns black and white and the pose of the sample photo is extracted. Now the pose is not repeated, but a similar one is generated. How to do it?

How am I supposed to use it? is it just I2I? I just loaded a bunch of big black empty nodes.

after a comfyui update i cant run anymore this workflow?? someone can update??

so month ago the florence, stops working, now i none is working

Sorry but i'm confused, the download is just 4 images, 4 pngs