Video Enhance AI update slow processing, want to return to the last version

I have been enhancing standard definitions video to HD with AI Video Enhance on my workstation PC. All has been going well. Then I updated to the latest release. Now, what was taking roughly 2 hours and 30 minutes is now taking 27 hours. Nothing has changed on my PC. Nor have the settings I’m using with the product. The only change was the update. How can I return to the previous release…which worked.
I have hundreds of videos to enhance for archival purposes and all work has now stopped. Any thoughts?
Larry Hatteberg

You can download the earlier releases from this forum however I would suggest you raise a support request at the main website as there haven’t been any reports like this…

Thanks, I’m rechecking my system. I appreciate your response!

I had that issue when switching from 1.3.8 to 1.7.1.
My encode speed went from 1.12spf to 11.24spf, it turns out that the model it was choosing was CPU only. For some reason, even though I had GPU (1) 1080ti selected it was using only CPU.
You can verify by selecting CPU render and comparing speeds.

Here is a collection of installers and release notes for Topaz Video Enhance AI. Not all versions are available:

last version with Artemis-MQ model (considered by some to be the best for DVD film content)

1 Like

Bookmarking the links. The new version of Video Enhance does not work for me at all. Depending on which output I choose, either nothing happens, or the program crashes. I’m going back to the previous version until the bugs are worked out.

nice job, videogeek. thanks.
is it possible to use the artemisMQ model with the newest version of the software somehow?

I did take a look and later versions use a different method of model loading. In 1.2.0 there’s a json file with the model and also a .dll file. 1.6.1 uses just a json and the model file. The json data is very different (it’s a text file, you can peek at them). 1.7.1 loads models from your Windows profile now. It’s not something one can (easily) hack. If one were a developer for the program, it could be done.

If it was my company, I’d bring back the Artemis-MQ 1.2.0 (as many users like it), and recall 1.7.0 & 1.7.1 due to the issues some have posted. Someone at Topaz decided to release 1.7.0 even with all its issues. I can’t understand why they would do that. Maybe having AMD and new Nvidia GPUs support to drive more sales was more important than output quality, stability, and performance. The timing of developer Matt Lanthrop’s leaving is suspicious. As if he knew the new version had serious issues, had a falling out with management over it and left (all pure speculation on my part).

I did investigate making VEAI 1.2.0 a portable app. I was able to get 1.2.0 to load, but it hung on the File Open dialog. I guessed that 1.2.0 and 1.6.1 only make registry changes, which makes it possible to export the two branches from HKLM and HKCU for each version and merge from the backups to the registry before running different versions. I’m sure there’s a portable app maker that can handle that, I just haven’t looked hard enough.

For now, doing it manually seems to work ok. I could write a loader with Autoit that would automate it, but it’s a very low priority project. 1.7.1 may be somewhat different since it’s saving models to your profile. Which may not matter (v1.8 may change that). Unique paths need to be used when installing each version. The registry needs to be correct for each version. They all share the same name for the registry trees, and that’s where the challenge lies.

Maybe someone with more experience with making non-portable apps portable can chime in.

In my own apps, I don’t use the registry, instead writing config data to .ini or .cfg file (old school and Linux style), and thus they are all portable. I wish more developers did that, as it gives the power users the opportunity to easily run multiple versions.