[Video] Simple Video Upscaler using Model (Patreon)

Hey Guys !! Here are Simple Video Upscaler Workflows:

https://drive.google.com/drive/folders/1-g8GjBdoJYlOsQI1MHZi1fJrGV7VAnmA?usp=sharing

• 1_0) Video2Video Upscaler
  

• 1_1) Video2Image Upscaler.
  

Both work on the same principle of Upscaling using a Upscaling Model. (Ersgan, Sformer... etc), No ksamplers are used in this workflow. You can Upscale Videos 2x,4x or even 8x times.

1_0) Video2Video Upscaler

It's a Video to Video Upscaling workflow ideal for 360p to 720p videos, which are under 1 min of duration.

Above than 1 min may lead to Out of memory errors as all the frames are cached into memory while saving.

1) Inputs

• Single Video Path - Right Click on the video and click "Copy as Path" and then paste the path in the Single Video Path Node
  

• Folder Input - Unmute the Nodes and Connect the reroute node to the Connect Path. Copy and Paste the Folder directory of the videos Folder. Select the video using the Selector Node.
  

2) Outputs

• You can leave the connection as it is, it will automatically save the output in the same Original Directory of the source video.
  

• Mute and use the use the Manual Path if you want.
  

DO NOT MUTE THE "Auto Output to Same Folder nodes group."

3) Video Settings

• Batch Range: Set the Batch Range - How much your PC can render at a time. Default is 24. Use Lower around 5 when doing a 720p to 4k conversion, To prevent going out of memory
  

• Video Settings: Set Fame Load Cap size to 24 if you want to test out different models.
  

• ***[MOVED to SAVINGS GROUP] FPS - Set FPS of the Output Video.
  

Do not downscale using force size, it introduces dicey pixels after upscale.

4) Upscaling

• 4x Upscale Model - Choose from a variety of 1x,2x,4x or 8x model from the https://openmodeldb.info Website. Remember that 2x, 4x, 8x means it will upscale the original resolution x2, x4, x8 times.
  

• Put the models here: ComfyUI\models\upscale_models
  

• 1x Refiner Model - You can use the 1x models here for refining the video first. If you don't want refinement, muting the node will give error, please reroute it manually.
  

5) Saving

• To Preview the upscale video please view from inside the explorer, Video combine node's preview don't display at full quality.
  

• The video will be automatically Moved to the path set above in Group "2) Output"
  

• Set crf to 20 to reduce the file size. CRF = Compression Percent.
  

• IMPORTANT! If you want to save longer videos or larger videos, please use the video2image Upscale workflow, the video combine node will crash and will run out of Memory soon.
  

• Upscalers workflows Link: https://drive.google.com/drive/folders/1-g8GjBdoJYlOsQI1MHZi1fJrGV7VAnmA?usp=sharing
  

• *Overwrite function is little buggy at the moment. Backup (Rename) if you are using the workflow for same video.
  

_______________________________________________________________________

1_1) Video2Image Upscaler

 It's a Video to Image Upscaling workflow ideal for 720p to 1080p videos which can be of any length.

As it will run in batches and render frames after every run, it will not go into out of memory.

Warning: It may produce 20 - 30 mbs size file for every frame so you can run out of storage space pretty fast.... use carefully.

Every setting is same as the 1_0) Vid2vid workflow above just the video settings is different

• Set the Lap Counter to "Increment" to enable auto skipping feature else you won't progress.
  

• Go to Side Menu > Extra Options > Auto Queue (Changed) then Queue Prompt to render all your video frames.
  

Make Sure to Put 0 in the Lap Counter before Starting for Final Render.

It Will require you to run via Auto Queue feature unlike 1_0) Vid2vid workflow which will run with just one queue. Please Watch this Post's Thumbnail video for more guidance.

__________________________________________________________________________

Custom Nodes:

https://github.com/daxcay/ComfyUI-JDCN

https://github.com/aria1th/ComfyUI-LogicUtils

Downloading Models

1) https://openmodeldb.info/

2) https://huggingface.co/uwg/upscaler/tree/main/ESRGAN

3) https://civitai.com/models/116225/4x-ultrasharp

or any site which provide the ESRGAN, SwinIR, HAT... types of model in .pth format

__________________________________________________________________________

Unsupported Models

See the Architecture type in the description of the Models:

Some of them are:

• SPAN

• Former

• RGT ... etc

which are not supported in comfyui and will give error:





__________________________________________________________________________

Some Samples for comparison are attached below:

__________________________________________________________________________

😁Easter Egg 🤪

Creating Big Text and Floating Emojis in Comfyui:

See end of the post's video for video tutorial

__________________________________________________________________________

Some new things are cooking... Will Come back soon.

Thank you all for your support.

• Discord Server : https://discord.gg/z9rgJyfPWJ
  
  - Jerry Davos
  
  <3333