1minute 8K Upscale

This is a workflow that quickly Upscale images to 8K resolution; simply drag and drop your image and click to run. The red section contains parameters that can be adjusted according to your needs.
Thanks for ControlAltAI in youtube and Kijai in github. This workflow is bulit on top of them and I learned a lot form their work.
Apisr in comfyui : github.com/ZHO-ZHO-ZHO/ComfyU...
Supir workflow:openart.ai/workflows/seven947...

[@101worlds]

So perfect! Thank you

[@user-pz2co7wg2j]

Apisr＆Supirは、Topaz Gigapixelよりも高性能ですか？
Is Apisr & Supir more powerful than Topaz Gigapixel?

[@genAIration]

Thak you!

[@xyzxyz324]

well explained! 4x upscale in my configuration in 15 min (rtx 3090 + 128 ram) --> 1792 x 960 px to (7168 x 3840 px) - 27.5 megapixels

[@matthewfuller9760]

how much ram did you actually use/ need for that?

[@Darksider-ct7qu]

Any solutions for Mac M1 pro ?

[@runebinder]

Been using this for a few days and have incorporated it into a workflow that does Txt2Img and InstantID first and it's great. Have yet to run into a VRAM issue yet, thank you :)

[@matthewfuller9760]

what size are you upscaling to?I am guessing you have 24 gb of vram?

[@runebinder]

@@matthewfuller9760, I wish lol. Gott a 2080Ti with 11GB VRAM. Can reliably get around 3000 x 4000 from this, although I have dropped that down of late to 2276 x 2926 as I tend to generate a lot of portrait and found the face difficult to detail after the upscale on higher resolution. All the SUPIR is doing is a x2 resolution, you don't need 24GB for that.

[@TaskSwitcherify]

LOL - "Teeth in mouth".

[@canvapikltd6305]

😁

[@manolomaru]

😵😵😵👍

[@AlexanderGarzon]

1 minute... more like 30minutes (4090 RTX, 64GRAM DDR5, AMD Ryzen 9 7900X3D and clean comfyui)

[@albert93911]

Takes me 2 minutes flat to generate an image, run it through a few ip-adapters,a reactor node, supir to 2k and ultimate upscale to 8K. On a 3090.

[@runebinder]

I've added it to a workflow with Txt2Img, InstantID and Face Detailer and that still takes less than half an hour on my system and I've only got a 2080Ti.

[@xyzxyz324]

sure that you are using the Lightning version of the model with required steps ?

[@Lost_Dawn]

More like 3 1/2 hours for me on my POS Intel Laptop. Never buy from Walmart... And that's on Final2X since I can't Run many other image upscalers.

[@siddheshd9]

nah would rather use topaz

[@trancaohuu2318]

I can't install Supir
Traceback (most recent call last):
File "E:\Data\ComfyUI-portable\ComfyUI
odes.py", line 1864, in load_custom_node
module_spec.loader.exec_module(module)
File "", line 940, in exec_module
File "", line 241, in _call_with_frames_removed
File "E:\Data\ComfyUI-portable\ComfyUI\custom_nodes\ComfyUI-SUPIR\__init__.py", line 1, in
from .nodes import SUPIR_Upscale
File "E:\Data\ComfyUI-portable\ComfyUI\custom_nodes\ComfyUI-SUPIR
odes.py", line 11, in
from .sgm.util import instantiate_from_config
File "E:\Data\ComfyUI-portable\ComfyUI\custom_nodes\ComfyUI-SUPIR\sgm\__init__.py", line 1, in
from .models import AutoencodingEngine, DiffusionEngine
File "E:\Data\ComfyUI-portable\ComfyUI\custom_nodes\ComfyUI-SUPIR\sgm\models\__init__.py", line 1, in
from .autoencoder import AutoencodingEngine
File "E:\Data\ComfyUI-portable\ComfyUI\custom_nodes\ComfyUI-SUPIR\sgm\models\autoencoder.py", line 6, in
import pytorch_lightning as pl
File "E:\Data\ComfyUI-portable\python_embeded\Lib\site-packages\pytorch_lightning\__init__.py", line 20, in
from pytorch_lightning import metrics # noqa: E402
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\Data\ComfyUI-portable\python_embeded\Lib\site-packages\pytorch_lightning\metrics\__init__.py", line 15, in
from pytorch_lightning.metrics.classification import ( # noqa: F401
File "E:\Data\ComfyUI-portable\python_embeded\Lib\site-packages\pytorch_lightning\metrics\classification\__init__.py", line 14, in
from pytorch_lightning.metrics.classification.accuracy import Accuracy # noqa: F401
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\Data\ComfyUI-portable\python_embeded\Lib\site-packages\pytorch_lightning\metrics\classification\accuracy.py", line 18, in
from pytorch_lightning.metrics.utils import deprecated_metrics, void
File "E:\Data\ComfyUI-portable\python_embeded\Lib\site-packages\pytorch_lightning\metrics\utils.py", line 22, in
from torchmetrics.utilities.data import get_num_classes as _get_num_classes
ImportError: cannot import name 'get_num_classes' from 'torchmetrics.utilities.data' (E:\Data\ComfyUI-portable\python_embeded\Lib\site-packages\torchmetrics\utilities\data.py)
Cannot import E:\Data\ComfyUI-portable\ComfyUI\custom_nodes\ComfyUI-SUPIR module for custom nodes: cannot import name 'get_num_classes' from 'torchmetrics.utilities.data' (E:\Data\ComfyUI-portable\python_embeded\Lib\site-packages\torchmetrics\utilities\data.py)
Can you help me ?

[@flyingdogsoft]

Supir is non-commercial only so what is the point of this video? research?

[@apache937]

use it anyways no one cares

[@CPDCT]

@@apache937 lol

[@apache937]

did zendaya consent

[@saymyname4325]

Does it work with sd15... I have only 6gb vram cant actually run sdxl 🥲

[@anonanonymous3781]

You can do it even with 4 GB VRAM, however you need Tiled VAE or else you'll encounter low RAM CUDA error. Just combine tutorial or upscaler you like with it. But I see that SUPIR has "tiled_vae decoder" option, os it should work.