Revolution in Image Enhancement: Discover AI's Free Upscaling Secrets! - ComfyUI Workflow

In this video, I reveal a revolutionary ComfyUI workflow that I have developed, which allows for upscaling images while adding incredible details. Using StableDiffusion XL Turbo, some controlnet and IPAdapter.

👍Don't forget to subscribe for more innovative tutorials like this one and leave a like if you find this video useful. Your support means a lot!

💬 Post your questions in the comments below. I'm here to help!

🔔 Turn on notifications so you don't miss our future videos on revolutionary technologies and image enhancement tips.

Settings:

You can adjust its strength to add more or less detail (range: 0.0-1.0) or completely bypass it.

Higher settings result in more hallucination. Recommended range: 0.15-0.4.

Adjust the Tile Size based on your VRAM capacity.

Upscaler Model - 4xNMKD:
Highly effective for upscaling.

Upscale recommended settings:
Range 0.5 - 1 for the top setting (multiplied by the upscaler ratio, e.g., using 4x_NMKD with 0.5 results in x2)
Range 1 - 2 for the bottom setting

IPAdapter:
Located in the middle column, it's deactivated by default. Activate it when aiming to achieve a specific style.

Prompt Adjustment:
Consider altering the prompt if the default one doesn't yield desired results. Adapt it to fit your needs.

ControlNet Depth
Sometimes, the hallucination is a bit high and we need a way to control the structure, You can activate it and play with it strength (around 0.4 is a good start)

[pablo.montero]

Just tested v4 and it's IMPRESIVE. Thank you for this contribution!!!

[joshualloyd]

I loved the content and really enjoyed the video. I can only speak for myself and please don't take this negatively. The music sounds good, but its personally very distracting. Thank you so much for taking the time to create this content for us.

[kidsword]

Hei bro.... Which is the link to download this model on my pc... I don't know 😊

[endersaka]

I don't find the workflow in the linked repository.

[WhySoBroke]

Great workflow!! Where would be the right place in the workflow to insert a subject specific Lora? To keep the subject consistent for example

[giusparsifal]

Hello and thanks for the video! Do it exist a easier wotkflow, this one is huge for a newbie as I am :), just to make face skin more realistic and less plastic? Just an img2img to add detail and do a skin more realistic :)

[alphakanalya]

I wanted to tune up my own alternative to web ehancers in comfy ui, but find your video and now i dont need it. Thank you, thas great!

[roxmay]

I can't find the upscale / Enhancement workflow in your github. Which one is it please ?

[aZiDtrip]

failed to install both controlnet auz preprocessors and was node suite.. found others with the same problem, some got it to work by reinstalling the opencv dependecies. but that didnt solve my problem.. any tips for me to get these nodes working ?

[bordignonjunior]

could you, please, share the json file with the workflow?

[Sethiris]

Wow is a great workflow! but I can't make it works, importing the image, I get soo missing nodes:
When loading the graph, the following node types were not found:
Text Concatenate
LoadImagesFromDir //Inspire
CR Prompt Text
CR Float To Integer
Int to float
Multiply
UltimateSDUpscale
Get image size
StringToNumber
Float Input [Dream]
Nodes that have failed to load will show as red on the graph.
do you have any links to download its?

[daniellimia]

For some reason, even bumping the Lora to 3.0, I don't seem the get much detail. I'm using the right one. Do you know why? Thanks! Btw, great workflow!

[maxziebell]

I am on a mac so I tried to get this to run on RunDiffusion. I failed as I only found a PNG in the GitHub repository. Is there a Comfy UI JSON for the upscaler setup for download or do we need to somehow rebuild it? Best regards Max

[hidalgoserra]

For some reasons for me doesn't work, everything it's in place I don't see missing nodes i receive this error in the terminal:
got prompt

ERROR:root:Failed to validate prompt for output 180:
ERROR:root:* Text Concatenate 63:
ERROR:root: - Value not in list: clean_whitespace: '' not in ['true', 'false']
ERROR:root:* Text Concatenate 66:
ERROR:root: - Value not in list: clean_whitespace: '' not in ['true', 'false']
ERROR:root:Output will be ignored
ERROR:root:Failed to validate prompt for output 237:
ERROR:root:Output will be ignored
Prompt executor has been patched by Job Iterator!
Prompt executed in 4.13 seconds
and i see read circle around the text concatenate nodes (both)

[masekin]

can you make a colab for this? i don't know how to

[edber]

This is great, would it be possible to do this to video smoothly by batching image sequences from a source video?

[jorgesarria]

I don´t know why when I upscale it makes a 4x4 tile image

[alecubudulecu]

Amazing. T❤Hank you for the details and explanation!

[haroldrandolf]

Hi could you please tell me how to fix this:

Prompt outputs failed validation
Text Concatenate:
- Value not in list: clean_whitespace: '' not in ['true', 'false']
Text Concatenate:
- Value not in list: clean_whitespace: '' not in ['true', 'false']

Error occurred when executing LoadImagesFromDir //Inspire:

Directory cannot be found.'

File "E:\zComfyUI\ComfyUI\execution.py", line 154, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)

File "E:\zComfyUI\ComfyUI\execution.py", line 84, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)

File "E:\zComfyUI\ComfyUI\execution.py", line 77, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))

File "E:\zComfyUI\ComfyUI\custom_nodes\ComfyUI-Inspire-Pack\inspire\image_util.py", line 30, in load_images
raise FileNotFoundError(f"Directory '{directory} cannot be found.'")


Error occurred when executing CLIPTextEncodeSDXL:

'g'

File "E:\zComfyUI\ComfyUI\execution.py", line 154, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)

File "E:\zComfyUI\ComfyUI\execution.py", line 84, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)

File "E:\zComfyUI\ComfyUI\execution.py", line 77, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))

File "E:\zComfyUI\ComfyUI\comfy_extras\nodes_clip_sdxl.py", line 44, in encode
if len(tokens["l"]) != len(tokens["g"]):

[Goldsmith.]

it will be possible with automatic1111?