Skip to main content
T
Tyler32669991g6q7
Participant
October 2, 2023
Question

Quality Loss / Compression on Linked Node / Graph

  • October 2, 2023
  • 2 replies
  • 269 views

Hello! I’ve been trying to troubleshoot an issue and I was hoping for some suggestions to try and fix this issue. I built a package that accomplishes several QoL functions. Inside the package/graph it functions perfectly. When linked to any other graph, including its duplicate, quality loss causes several functions to cease working. Pictures included show some examples of this loss comparing outputs in original graph and to linked results. Top two are the results after being linked while bottom two are the original results. The section I have for scanning dark values via histogram scans tends to be dropped after being linked- and the section I have that introduces samples of colored noise tends to blur towards singular color values.

The graph uses only atomic nodes, heavily reliant on scans and blends. I’ve tried changing the graph properties output format and resolution to absolute, adding transform nodes after inputs to inherit parent properties, and upping bit depth to double the graphs original value for the scan outputs but data is still lost or heavily compressed down the chain. I’m still actively troubleshooting but am at a loss with this. Any suggestions would be great!

This topic has been closed for replies.

2 replies

Cyril Dellenbach
Community Manager
Cyril DellenbachCommunity Manager
Community Manager
October 3, 2023

Hi Tyler,

 

Thanks for the message and the details.

 

Considering the Switch vs Blend issue, I'm suspecting a problem with the cache, preventing data to load properly. Would you mind sharing your graph, so we can investigate on this potential bug?

 

Make sure to export the graph with its dependencies (RMB on the package > export with dependencies...) and you can send it to cdellenbach@adobe.com

 

Best regards,

 

Cyril Dellenbach (Micro) | QA Support Artist | Adobe
T
Tyler32669991g6q7Author
Participant
October 6, 2023

Hello! Sorry I didn't see the message- I'll upload the graph / recreate the issue sometime tomorrow and email it over.


Best,

 

Tyler S.

    T
    Tyler32669991g6q7Author
    Participant
    October 2, 2023

    I was able to isolate a part of the graph regarding the scans. After the data goes through all of my grayscale switches the output fails to return properly. If the scan is done at the start then it returns expected results. I do have consecutive switches in my graph to dictate whether or not data travels down certain paths, though admittedly I'm not sure if this actually prevents substance from doing so. Is it best to just remove these switches? Is there some interaction with these types of switch chains and irregularities when linking?

    T
    Tyler32669991g6q7Author
    Participant
    October 2, 2023

    When replacing the grayscale switches with copy blends in the isolated graph, the outputs work as expected. I will employ this change in the full graph and update if this solves the issue. There must be some reason for this node to causing issues when it comes to linking or referencing graphs.

    T
    Tyler32669991g6q7Author
    Participant
    October 2, 2023
    Replacing partially worked. For some reason when referencing the graph the bool logic would flip and that ended up being the other half of the fix. Not sure why this was the case, but even though the original graph doesn't function as desired the referenced one does, which is the end goal.
    Terms & ConditionsAccessibility statement
    Using the community Terms of use Privacy Cookie preferences Do not sell or share my personal information ad choicesAdChoices