you can test steps here to delete the daclip model and rerun the installer, and if any issues do remain, send an email to support@topazlabs.com with your app logs and we can have a look.
@TPX - I can assure you we donât have any AI agents that reply at Topaz on forums or emails. Bin.sun is a Researcher here at Topaz Labs 
5 Likes
Just been trying out Wonder 3 for painted comic book style art and it is adding noise to the image (patches of coarse grain and patterns like it is imitating bad scanning artifacts). Of course, it almost certainly is more optimised for photos, and understandably so. But considering the speed difference, I hope Wonder 1 will remain available into the future as it is fast and gives excellent results with cleaning up that particular art style (especially cleaning up outlines), a much more natural result in many cases to the standard Art model. The only quibble is that Wonder 1 slightly reduces dynamic range by lowering highlights a little (correctable in post).
If you can share here, the Original, as well as Wonder 1 results vs Wonder 3 unwanted results we can relay to the development team. Here or at support@topazlabs.com and we can relay with the development team working on Wonder 3.
1 Like
I gave another go at the Wonder 3 model, cloud, and itâs great.
For such smartphone photos shot inside with good light, it does great on faces and hairs. But it mess up the text, add new noise and sometimes changes a bit the shape of small symbols.
2 Likes
TL;DR Looks like if you use Norton Security on W11 (Home, 25H2, fully patched) then install will need \program data\Topaz Labs\Topaz Photo\ whitelisted for anti-virus scanning. Perhaps worth putting somewhere less like this rabbit hole to save other users hours (literally) of repeated installation attempts.
On my second attempt at installing 1.5.1 (this time from 1.4.0, I think: showed 2 updates available when I got it to run), the install stalled reporting a problem with installing the wonder model. If there was any text to show which, this wouldnât fit in the message box. I let the install run on, then - while the neurosever deps were downloading - went into Norton Security, Advanced Security, Antivirus, Exclusions, and added the \Program Data\etc to my list of exemptions.
On running Topaz Photo at the end of installation, and loading the first image, Photo had to download a couple of the model files. Encouragingly, it seems as though each model that has failed to install will only have to be downloaded once.
HTH
2 Likes
Thanks Mike, appreciate this and will relay to Lingyu and the Support team
1 Like
Preserve Text seems to help a bit with reducing text changes, but itâs not perfect.
We will need your app logs to support@topazlabs.com (From Topaz Photo as a standalone, then Help menu, Open Logs folder) and we can see what is happening
1 Like
A post was split to a new topic: Discussion | Denoise AI in Capture One Beta
For some reason, with WOnder 3, the first time I use it on an image it works fine, but reapply it with same settings reloading same image, it now suddenly erases/blurs out a bunch of detail in certain areas. Not sure its my pc spazzing out or models beign corrupot on my pc?
I have just email you and attach the logs for that event. Thank you.
1 Like
Exceptionally buggy on my end (RTX 4060) - seems to take forever for Denoise Max in particular (progress bar does not move) and Wonder 3 speed is unstable (way different processing times for similar images) - just me or what?
2 posts were split to a new topic: Dicsussion | Denoise Max Youtube reactions and comparisons
Denoise Max status bar is a known issue and will be worked on in future updates, as well as optimizations for it and Wonder 3, Super Focus 3.
Keep in mind that the first file of a session will be slower as the neuroserver is invoked, then subsequent files of the same session.
3 Likes
Does a session mean a cold boot of the applicationâs executable? So every single cold boot will invoke the NeuroServer?
When the tool itself is invoked. If any issues occur, send logs to support@topazlabs.com with details and we can have a look to see if any errors occur in the processing itself. Ensure to also send the images being edited (originals) and we can test to see if can replicate.
A post was split to a new topic: Topaz Video - Starlight HQ bug
