Auto Crop -> CCSR -> SUPIR -> UltimateSDupscale -> High quality upscale
5.0
5 reviews
Description
This workflow is not meant to be fast, but aim for a VERY HIGH quality output. It outperform anything I know about upscale. Better than Topaz. Better than Magnific. But it have a down side, it is heavy on hardware when using the full workflow with large output size.
👉What it do:
It upscale a poor image into something mind-blowing while conserving integrity of the picture.
👉How it work:
The workflow look very big at first but is very simple to use. You can start by checking if you have all models in the workflow and can by tested directly. If you have problem, you can try to activate 1 group node at the time and see how it goes before going to next step.
1- We crop the image at a SDXL most closest size. This is done with my custom nodes attached in files. Copy FRED_AutoCropImage_SDXL_Ratio.py inside your \ComfyUI\custom_nodes.
2- We first upscale the image and clean it before it reach SUPIR. This is very important as SUPIR work best with a clean image, and simple upscale with model or upscale by node doesn't work well all the time. To do that, I use a CCSR upscale method. That upscale the image and keep it integrity and clean it. This mostly work good for noisy picture, with lots of artifacts, but is very hardware dependant and take sometime to load model for the first run.
3- If the CCSR upscale didn't worked to clean the picture, we can use StableSR group. Usually not necessary.
4- When the image is enough big (around 2x a SDXL standard), SUPIR will do it's magic. Adding skin texture and imperfections, adding detail to a landscape. It have its flaw, like adding hair where there isn't. To work around, we lower the CFGs but the common settings I finely found work very good.
5- Don't worry if there was too much skin imperfections as they will be finetune in ultimateSDupscale. I do another 2x here and using realvisXL30Turbo with the upscaler 4xNMKD-Superscale0SP_178000_G it add skin detail at a micro level.
6- Saving the picture in 2 different format, 1 with only a watermark at the bottom, and 1 with a compare of the original with automatic text placement. There is also a BEEP sound when the process end.
👉Where it work the best:
This workflow work so good for portrait image of humain on image around 512x512. So old SD1.5 image here we are for the upscale.
It work great on other pictures too, some tweeking maybe required.
👉More infos in workflow.
This workflow contains around 250 nodes but you can all find them in comfyui_manager except for my custom node that is attached in a zip file. Put it in folder mentionned earlier. You need numpy, cv2, PIL and Torch to run, but most of you will all have that already. If not, then do a pip install numpy, and the other one.
It also take a lot of ressouces but the result is worth it. For a normal upscale, it take around 8min/image upscaled on my RTX4090.
🔥Help me improving this workflow. If you find a better way to parse data or a better model, please share.
🔥Most install problem is not due to the workflow by itself but missing dependency or missing custom nodes or models. Be sur you're all set before posting a bug in the comments.
PaintShop was used ONLY to convert image to jpg as the result is around 33mb. No modification was done within that tool.
I hope you gonna have fun as much as I have to upscale pictures using this workflow.
Node Diagram
Discussion
When loading the graph, the following node types were not found:
- Get resolution [Crystools] 🔗
- Primitive integer [Crystools] 🔗
- Switch any [Crystools] 🔗
- Primitive float [Crystools] 🔗
- Switch image [Crystools] 🔗
- ApplyStableSRUpscaler 🔗
- FRED_AutoCropImage_SDXL_Ratio 🔗Nodes that have failed to load will show as red on the graph.
Please, can you say what I am doing wrong? I downloaded missing nodes, but this message appeared
You need to install crystools, follow it's installation process on ( https://github.com/crystian/ComfyUI-Crystools?tab=readme-ov-file#installation ),
StableSR checkpoint webui_768_v139.ckpt missing (https://huggingface.co/Iceclear/StableSR/blob/main/webui_768v_139.ckpt ).
And FRED_AutoCropImage_SDXL_Ratio.py (in the zip in the file of this web page), as stated in note have to be copyed in the comfyui/custom_nodes directly then restart comfyui.
FRED_AutoCropImage_SDXL_Ratio 🔗 Where can I download this?
on this site on the top. I think you must be in a normal browser to see it.
RuntimeError: Input type (struct c10::Half) and bias type (float) should be the same
what node are in red in graph for that error?
Single Image or Batch?
⌨️ CR Load Image List
It gets stuck in these two places
for now, just send image directly to the colapsed context and remove the batch node the seed and the browser for folder image or bypass them
Thank you . I bypass them
When loading the graph, the following node types were not found:
- ApplyStableSRUpscaler 🔗
https://github.com/WSJUSA/Comfyui-StableSR
Even though it was installed, it still doesn't recognize it. Do you know why ?
try another workflow that use stableSR to see if it also not working.
I dont know how to install this node :(
You need to install the models in the StableSR folder in models/stablesr
and in the checkpoint folder. Following the instructions found here: https://openart.ai/workflows/mad4bbq/stablesr-comfyui-fixedultimatesdupscale/6kTHKAKhlMH9g09lfXG0
Do we really need to have this model?
real-world_ccsr.skpt
yes
if you want to use CCSR upscale for low quality pictures input.
https://github.com/kijai/ComfyUI-CCSR
Another question, why can't I move any box? How do I release them?
Because in the last part it means 9. In the Original resolution section, it says that no photo is attached
right click the node then "pin". i do that to avoid accidently move them around
RTX 4090, but it takes me 30 minutes to upscale each time, the image size is 896*1125, and the parameters are all the default parameters
Somthing wrong in your config, it cannot take that long. I do have an RTX4090 too, and the first run, when no models is loaded is longer but after it take around 5-6 min each.
Registor to say THANK YOU SO MUCH!
You are a legend!
(Edited)The result is amazing! but there are some artifact around the face, any idea?
Here is my image: https://postimg.cc/JsvPCb4D
Hello, I think you got that bad result because you use the fp16 supir model, try using the SUPIR_v0Q. You can also add some words to the positive prompt in the supir group. Like: 30 year old man with black beard and stripes sweater.
But if you can, please post me the original picture, i'll try it with the version 7 of the workflow, and to my new version.
I made it, prepare yourself psychologically before you look at the result ;). Zoom on the picture...
(Edited)OH MY GOD!!!!!! That's incredible!
Yes! I'm using the fp16 version of SUPIR. I'll try the original version, Thanks!
Yes! Here is the original image: https://postimg.cc/yk9RdGHP It's just 100x150, I never thought it could be upscaled to have such amazing detail.
(Edited)Mind blowing!!! 🤯
Nope, The checkpoint doesn't seem to be the key reason. Here is the result I got wth SUPIR_v0Q: https://postimg.cc/F7vwR7dn .
Some part of my picture seems to be blurry.
pass your picture with only crop then ccsr. take that picture put it back in the input then only do StableSR -> supir -> ultimateSDupscale. It should do it
It's good: https://postimg.cc/0KGSn6WH But not as good as yours :(
I finally made it, it turns out to be a silly mistake: I'm using wowxl420 for UltimateSDUpscaler but using 10 steps as your default configuration.
realvis give more realistic skin and less deformation for a higher denoise. If you find a way to get NSFW image working in SUPIR with a model without deformation, please tell
Error occurred when executing KSampler:
Allocation on device 0 would exceed allowed memory. (out of memory)
Currently allocated : 2.43 GiB
Requested : 29.79 GiB
Device limit : 22.49 GiB
Free (according to CUDA): 0 bytes
PyTorch limit (set by user-supplied memory fraction)
: 17179869184.00 GiB
File "S:\ComfyUI_windows_portable\ComfyUI\execution.py", line 151, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "S:\ComfyUI_windows_portable\ComfyUI\execution.py", line 81, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "S:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-0246\utils.py", line 381, in new_func
res_value = old_func(*final_args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "S:\ComfyUI_windows_portable\ComfyUI\execution.py", line 74, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "S:\ComfyUI_windows_portable\ComfyUI\nodes.py", line 1344, in sample
return common_ksampler(model, seed, steps, cfg, sampler_name, scheduler, positive, negative, latent_image, denoise=denoise)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "S:\ComfyUI_windows_portable\ComfyUI\nodes.py", line 1314, in common_ksampler
samples = comfy.sample.sample(model, noise, steps, cfg, sampler_name, scheduler, positive, negative, latent_image,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
is it possible to solve this problem?
i never faced that error. do you have a node with red surrounding color when it happen?
:1x\1xJPEGDestroyer.pth: up;Can you share this amplifier? Can't find it, thank you
the best workflow for restoring old photos. Other sites are fighting because of much worse results. I'm surprised you put it in the public domain. Thank you so much. You are brilliant.
:1x\1xNoiseDestructor.pth: Can't find it , please share
https://civitai.com/models/147838?modelVersionId=164924
Hi ! I am really delighted with this workflows !
In my opinion it's not lying when you say "Better than Topaz. Better than Magnific.".
However, I still encounter several difficulties and perhaps you could help me : EARLY UPDATE
1. I don't manage at time to make SRUpscaler working. I look at other comments, i tried to download every .chk on hugging face, but it always don't find the node. Can you provide a tips for install ? Can't from ComfyUI manager right ?
2. The SUPIR created picture don't jump automaticly to ultimate upscale, and so i have to save the SUPIR result and manualy bring it in load image of Upscaler. Is that only a settings of "external picture" node ? Did i miss something ?
Big thanks for your work !!!
in external load, put it to false and put a real valud folder in there even if not used as it will create error.
for stableSR follow closely the installation proposed on github. this node is not easy to install. for info, stablesr group in my workflow should only be used on very noisy picture other else it can produce worst result
Thanks for your answer !
Set a valid picture in external loader resolve the issue.
And i precise it for any possible readers, i needed to erase all previous install of stableSR and reinstall with git to solve it. And must admit it use is really more for very noisy picture.
A last assertion and question : i tried sevaral checkpoints, and my best results are, from far, with the big sized realcartoonXL v6. Results are clearest, intricate and not hallucinated.
Have you, while you use your workflow some much, a recommendation for good results checkpoints ? Of course it must depends on the desired goal, but i figured that some checkpoints works better, even on in styles that are not supposed to be theirs.
Once again, majors thanks for this workflow
(Edited)I have an important question. Why do we have two Load Images?
What is the use of the second Load Image?
Should it be the same in both photos?
For example, if a person is in the first part, should the same person be in the second part?
i'm not sure if i get your question right. you load a pictures to correct it. that's the purpose of this workflow. if you talking about the use external images, i added this just for convenience to be able to use only the ultimatesdupscale and a compare to the orginal image. but to work you must load a valid image or the node not working.
ApplyStableSRUpscaler problem. I've tried to install stableSR node and 2 models several times sucessfully. But I still always got this "ApplyStableSRUpscaler" error. What's wrong? Please help
Hi !
If i remember well, i had to erase previous install and follow this instruction :
Source: https://github.com/gameltb/comfyui-stablesr
The best way to install StableSR for ComfyUI in Windows is to go to your 'ComfyUI\custom_nodes' directory, open a CMD prompt, then type "git clone https://github.com/WSJUSA/Comfyui-StableSR.git" (no parenthesis). Then start ComfyUI.
Models required:
- Put the StableSR webui_786v_139.ckpt model into Comyfui/models/stablesr/
- Put the StableSR stablesr_768v_000139.ckpt model into Comyfui/models/checkpoints/
- Download the ckpt from HuggingFace https://huggingface.co/Iceclear/StableSR/
Good luck
Error occurred when executing "SUPIR_first_stage:
" maximum recursion depth exceeded while calling a Python object
Your image input is smaller than the tiling of the node where it failed.
Error occurred when executing SUPIR_first_stage: maximum recursion depth exceeded while calling a Python object File "/data/applications/ComfyUI/execution.py", line 151, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) File "/data/applications/ComfyUI/execution.py", line 81, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) File "/data/applications/ComfyUI/execution.py", line 74, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) File "/data/applications/ComfyUI/custom_nodes/ComfyUI-SUPIR/nodes_v2.py", line 335, in process decoded_images = SUPIR_VAE.decode(sample).float() File "/data/applications/ComfyUI/custom_nodes/ComfyUI-SUPIR/sgm/models/autoencoder.py", line 318, in decode dec = self.decoder(z, **decoder_kwargs) File "/data/python-venvs/comfyui/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/data/python-venvs/comfyui/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl return forward_call(*args, **kwargs) File "/data/applications/ComfyUI/custom_nodes/ComfyUI-SUPIR/SUPIR/utils/tilevae.py", line 702, in __call__ return self.net.original_forward(x) File "/data/applications/ComfyUI/custom_nodes/ComfyUI-SUPIR/SUPIR/utils/tilevae.py", line 702, in __call__ return self.net.original_forward(x) File "/data/applications/ComfyUI/custom_nodes/ComfyUI-SUPIR/SUPIR/utils/tilevae.py", line 702, in __call__ return self.net.original_forward(x) [Previous line repeated 481 more times] File "/data/applications/ComfyUI/custom_nodes/ComfyUI-SUPIR/SUPIR/utils/tilevae.py", line 706, in __call__ self.net.to(original_device) File "/data/python-venvs/comfyui/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1173, in to return self._apply(convert) File "/data/python-venvs/comfyui/lib/python3.10/site-packages/torch/nn/modules/module.py", line 779, in _apply module._apply(fn) File "/data/python-venvs/comfyui/lib/python3.10/site-packages/torch/nn/modules/module.py", line 779, in _apply module._apply(fn) File "/data/python-venvs/comfyui/lib/python3.10/site-packages/torch/nn/modules/module.py", line 779, in _apply module._apply(fn) [Previous line repeated 2 more times] File "/data/python-venvs/comfyui/lib/python3.10/site-packages/torch/nn/modules/module.py", line 803, in _apply with torch.no_grad(): File "/data/python-venvs/comfyui/lib/python3.10/site-packages/torch/autograd/grad_mode.py", line 81, in __enter__ torch.set_grad_enabled(False) File "/data/python-venvs/comfyui/lib/python3.10/site-packages/torch/autograd/grad_mode.py", line 184, in __init__ self.prev = torch.is_grad_enabled()
What is the size of the image before SUPIR? I have hard time overpass the double of SDXL normal size. So try to only put a value of 1.5 in the SCALE_FACTOR in the crop group. If it's not the problem, you may have to reinstall torch.
Almost there. Some CCSR_Upscsale error. Can someone help me with this?
Error occurred when executing CCSR_Upscale:
Attempting to load the storages to the PRIVATEUSEONE device but torch.storage._StorageBase.privateuseone() or torch.storage.TypedStorage.privateuseone() is not generated. Please use torch.utils.generate_methods_for_privateuse1_backend to generate storage.privateuseone() method first. File "C:\Users\1\ComfyUI\execution.py", line 151, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) File "C:\Users\1\ComfyUI\execution.py", line 81, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) File "C:\Users\1\ComfyUI\execution.py", line 74, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) File "D:\Anaconda\envs\comfyui\lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context return func(*args, **kwargs) File "C:\Users\1\ComfyUI\custom_nodes\ComfyUI-CCSR\nodes.py", line 72, in process empty_text_embed = torch.load(os.path.join(script_directory, "empty_text_embed.pt"), map_location=device) File "D:\Anaconda\envs\comfyui\lib\site-packages\torch\serialization.py", line 1026, in load return _load(opened_zipfile, File "D:\Anaconda\envs\comfyui\lib\site-packages\torch\serialization.py", line 1438, in _load result = unpickler.load() File "D:\Anaconda\envs\comfyui\lib\site-packages\torch\serialization.py", line 1408, in persistent_load typed_storage = load_tensor(dtype, nbytes, key, _maybe_decode_ascii(location)) File "D:\Anaconda\envs\comfyui\lib\site-packages\torch\serialization.py", line 1382, in load_tensor wrap_storage=restore_location(storage, location), File "D:\Anaconda\envs\comfyui\lib\site-packages\torch\serialization.py", line 1311, in restore_location return default_restore_location(storage, str(map_location)) File "D:\Anaconda\envs\comfyui\lib\site-packages\torch\serialization.py", line 391, in default_restore_location result = fn(storage, location) File "D:\Anaconda\envs\comfyui\lib\site-packages\torch\serialization.py", line 363, in _privateuse1_deserialize raise RuntimeError(f'Attempting to load the storages to the {backend_name.upper()} device '
The best I can tell is to go follow the installation instruction on CCSR github page.
The Node under Stable SR (Clean & Soften)-Checkpoint is undefined :(
At CCSR Upscale (for noisy and small pictures)
!!! Exception during processing!!! invalid load key, '\xf8'.
Traceback (most recent call last):
File "G:\ComfyUI\ComfyUI\execution.py", line 151, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\ComfyUI\ComfyUI\execution.py", line 81, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\ComfyUI\ComfyUI\execution.py", line 74, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\ComfyUI\python_embeded\Lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "G:\ComfyUI\ComfyUI\custom_nodes\ComfyUI-CCSR\nodes.py", line 78, in process
load_state_dict(self.model, torch.load(ccsr_model, map_location="cpu"), strict=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\ComfyUI\python_embeded\Lib\site-packages\torch\serialization.py", line 1040, in load
return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\ComfyUI\python_embeded\Lib\site-packages\torch\serialization.py", line 1262, in _legacy_load
magic_number = pickle_module.load(f, **pickle_load_args)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
_pickle.UnpicklingError: invalid load key, '\xf8'.
Your model look to try to load in CPU, you may have something wrong in your comfyui installation
I installed ComfyUI normally,not sure how it switches into my CPU
Is it possible to run on lower gear like 3070 8gb? On lower settings?
Yes and No. Yes if you lower the prescale factor to like 1 and put all tiling option to ON. That could take a long time but it's possible. But you'll never get same result as in the pictures linked here.
After the latest update, I encountered a problem and that is that
In the CCSR_Upscale section
2024-06-18 10:46:25,241- root:212- INFO- Loaded ViT-H-14 model config.
It does not go further and cannot load the model. It seems to be stuck at this point
do a cmd in the CCSR custom node, then do this command: git pull
if that doesn't work, be sure CCSR working properly, try their example workflow in the folder examples
What's the most efficient way to integrate creative upscale with this workflow?
Truly near-perfect upscaling by the way, best available by far and I don't see how it can get better, at this point it's all about reducing the resources/time required.
Hope you can guide me. I ran the workflow successfully, but the results were terrible. I was doing something wrong. Please help me thank you very much. https://drive.google.com/drive/folders/1LZ0lcApqoQXP9SMzQUZ4n-FvD03-fEr-?usp=sharing
Got an error here, please help.. Error occurred when executing CCSR_Upscale: 'NoneType' object has no attribute 'movedim' File "C:\Users\graef\ComfyUI_windows_portable\ComfyUI\execution.py", line 151, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\graef\ComfyUI_windows_portable\ComfyUI\execution.py", line 81, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\graef\ComfyUI_windows_portable\ComfyUI\execution.py", line 74, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\graef\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context return func(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\graef\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-CCSR\nodes.py", line 82, in process image, = ImageScaleBy.upscale(self, image, resize_method, scale_by) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\graef\ComfyUI_windows_portable\ComfyUI\nodes.py", line 1635, in upscale samples = image.movedim(-1,1) ^^^^^^^^^^^^^
Hi @Michael Knight, I'm using this pipeline for upscaling games assets: the face texture map. This pipeline does a great job for white person, but the result is really bad for black person. Any ideas?
Here is my test input if you are interested: https://i.postimg.cc/90NJ6MNC/face-227236-0-0-color.png
I don't know if my workflow will recognize your image, its not even a person, maybe if you add "black men" in supir prompt and lower the cfg for supir to 1.5 and 2 for the 2nd.
Hi, I downloaded and unzipped FRED_AutoCropImage_SDXL_Ratio into my Custom Nodes folder and restarted Comfy but it's not picking it up. The workflow says it's missing and in red, if I go into Manager > Install Custom Nodes and filter it to Import Failed that's clear. Any ideas?
Could the RetinaFacePM node which can also crop faces work as an alternative?
place the single file xxx.py in the custom node, do not create a folder
I found that Juggernaut-XL_v9_RunDiffusionPhoto_v2 better than wowxl_420 in my case, better cosistency and less hair.
Thanks for the info
Sorry if I sound frustrated, but I'm totally baffled why someone who puts this amount of effort into constructing, testing and freely sharing a complex workflow like this, managed to make it SO hard for people to puzzle together where they can find all the components (???)
I've been searching around for a couple of hours now, and HOPEFULLY I managed to scrape all the necessary sources together. (downloads still running, so...)
See the list below for the results of my searching (would even be better with the final locations for each file, but I'm too tired now and as I said not able yet to test anyway):
=======================================
The ComfyUI workflow itself:
On the page right here, "Download" button on the upper right
FRED_AutoCropImage_SDXL_Ratio:
(FRED_Auto_Crop_Image_SDXL_Face_Detect_v1.py)
On the page right here, as a ZIPfile via "resources" on the right
realvisxlV30Turbo_v30Bakedvae:
I can only find "V50"... realvisxlV50Turbo_v30Bakedvae, hopefully that will also be able to made to work:
https://civitai.com/models/139562?modelVersionId=798204
SUPIR-v0F:
https://huggingface.co/camenduru/SUPIR
https://huggingface.co/camenduru/SUPIR/tree/main
sai_xl_depth_256lora:
https://huggingface.co/lllyasviel/sd_control_collection
https://huggingface.co/lllyasviel/sd_control_collection/blob/7cf256327b341fedc82e00b0d7fb5481ad693210/sai_xl_depth_256lora.safetensors
stablesr_768v-000139:
webui_768v_139:
More info: https://openart.ai/workflows/mad4bbq/stablesr-comfyui-fixedultimatesdupscale/6kTHKAKhlMH9g09lfXG0
https://huggingface.co/Iceclear/StableSR
https://huggingface.co/Iceclear/StableSR/tree/main
4x_NMKD-Superscale-SP_178000_G:
(why link to a 17GB archive of ALL upscalers known to mankind???)
https://openmodeldb.info/
https://icedrive.net/s/43GNBihZyi
real-world_ccsr.skpt:
https://github.com/kijai/ComfyUI-CCSR?tab=readme-ov-file
https://drive.google.com/drive/folders/163x-ynZ2fYPxvo_DkTzhZy5W2jeSBM9S
wowXL420_V420:
https://tensor.art/models/708327119123643169
Hopefully this is of some help...
Hello, I have looked at your workflow and found it very shocking. The zoom in effect is simply perfect, but I am curious how you know that SUPIR needs to remove noise from the original image before zooming in? As far as I know, not many people have discovered this knowledge point, so I am really curious about how you discovered it? Through a large number of experiments? Or reading papers? If you have time, please make sure to let me know. Thank you very much!
ComfyUI Error Report
## Error
Details
**Node
ID:** 174
Type:**CCSR Model Select
**Node
**Exception Type:** urllib.error.URLError
*Exception Mesage:.* <urlopen error [SSL: CERTIFICATE VERIFY FAILED] certificate verify failed: self-signed certificate in certificate chain ( ss1.c:1007)>
## Stack Trace
Could you please tell me what is the reason for this?
If anyone runs into issues with StableSR module not loading. Go to your comfyUI/custom_nodes and run "git clone https://github.com/Arthurzhangsheng/Comfyui-StableSR.git ". I tried this same thing with another repo suggested and it did not work but this one should
Could this be better with flux?
Node Details
Primitive Nodes (51)
Any Switch (rgthree) (1)
ApplyStableSRUpscaler (1)
Context (rgthree) (4)
Context Big (rgthree) (5)
Display Any (rgthree) (1)
FRED_AutoCropImage_SDXL_Ratio (1)
Fast Groups Bypasser (rgthree) (1)
Get resolution [Crystools] (12)
GetNode (2)
Image Comparer (rgthree) (5)
Note (2)
Primitive float [Crystools] (5)
Primitive integer [Crystools] (4)
PrimitiveNode (1)
Reroute (1)
SetNode (2)
Switch any [Crystools] (2)
Switch image [Crystools] (1)
Custom Nodes (63)
- JDC_ResizeFactor (1)
- CR Simple Text Watermark (1)
- CR Multi Upscale Stack (2)
- CR Apply Multi Upscale (2)
- CR Image Panel (1)
- CR Image Border (1)
- CR Overlay Text (3)
- CR Load Image List (1)
- CR Batch Process Switch (1)
ComfyUI
- CheckpointLoaderSimple (1)
- VAEEncode (1)
- SaveImage (2)
- ControlNetApply (1)
- ControlNetLoader (1)
- PreviewImage (2)
- LoadImage (2)
- UpscaleModelLoader (1)
- CLIPTextEncodeSDXLRefiner (2)
- CLIPTextEncode (2)
- KSampler (1)
- VAEDecode (1)
- ImageResize+ (1)
ComfyUI_tinyterraNodes
- ttN text7BOX_concat (1)
- AV_ControlNetPreprocessor (1)
- CCSR_Model_Select (1)
- CCSR_Upscale (1)
- ResizeImageMixlab (1)
- TextInput_ (2)
- SUPIR_model_loader_v2 (1)
- SUPIR_conditioner (1)
- SUPIR_first_stage (1)
- SUPIR_decode (1)
- SUPIR_sample (1)
- ColorMatch (2)
- MathExpression|pysssss (8)
- CheckpointLoader|pysssss (2)
- PlaySound|pysssss (1)
- LoraLoader|pysssss (1)
- ShowText|pysssss (2)
- ReActorRestoreFace (1)
- Seed Generator (1)
- UltimateSDUpscale (1)
- JWImageResize (1)
Model Details
Checkpoints (3)
realvisxlV30Turbo_v30Bakedvae.safetensors
stablesr_768v_000139.ckpt
wowXL420_v420.safetensors
LoRAs (1)
DetailedEyes_V3.safetensors