My installation cannot load models. Video AI loads and runs, no problem, after I logged in but when it’s time to preview or export, VAI starts to load a model (“Loading Model 0s” and three animated dots in the output window) but no data is coming in from the internet. I can see from my network tools that the PC has opened a number of sessions but incoming data rate is 0 Kbps. The firewall’s disabled so it’s not that.
I can access the internet, like this forum, without issues on the same PC.
Exact same issue here. Needs urgent fixing.
How large are these model files supposed to be anyway? Proteus one appears as 32 MB.
Update: Proteus is the model that will not process for me. All others are working at much faster speeds. Of course, Proteus is the model I use most. Looking forward to the patch as this will speed up many of my projects. Thanks!
@suraj @eric @yazi.saradest @TPX @nipun.nath
UPDATE:
I confirm that, with my PC, only the “Gaia” and “Theia” models work (while the “Proteus” and “Artemis” models do not seem to be loaded properly, and thus the process stops when the AI models are loaded).
I can confirm, however, that my 3070 is finally (especially with “Gaia”) being utilised to the maximum (GPU-Z shows a constant 99% utilisation with the “Gaia” model, while with “Theia” we are at a steady 60% with more oscillations).
I can confirm, however, that you are on the right track: increased model quality coupled with better utilisation of modern PC hardware and thus increased speed!
Ran a quick test on my Mac Studio on macOS 13.1. Gaia and Theia work but Proteus and Artemis do not.
Sorry to keep adding information, but it may be helpful to know that I have the latest NVidia studio driver installed (528.02) released on 1/5/23.
Same here. RTX 3080 and it stucks at loading model.
For good measure, here are my logs:
2023-01-07-12-05-37-Out.zip (41.5 KB)
EDIT: Fun fact, TVAI compresses the logs for you into a .RAR file, but the forum only accepts the .ZIP kind. 
I have tested Video AI Beta 3.1.0.0.b on both my 16 in. M1 Max MacBook Pro and on my 12-core Late 2019 Mac Pro using S-VHS interlaced footage recorded at 720X480 and upscaled to FHD (1920X1080). I had no problems using the software on either machine, and I both cases I saw substantial improvement in the processing speed compared to the previous versions. In the case of the M1 Max MacBook Pro the speed increased from 0.09 to 0.05 sec/frame and in the case of the Mac Pro the speed increase was from 0.14 sec/frame to 0.09 sec/frame. This is using either the Dione Robust model or the Dione DV 2X model. Great job and kudos to the Topaz Video AI team for their hard work. I look forward to further beta updates. Tom
I can confirm a crazy uplift in performance, when a model loads - Gaia has now 14fps, previous stable 3.0.0.11 had it at around 5fps on that destroyed D&D cartoon I’m testing all VEAI3 versions with. All on RTX 3090 GPU (PCI-Ex 3.0), 7940X CPU (fairly old at this point) with 528.02 driver.
Proteus, Artemis and Dione models don’t start any processing as for now, but that’s fixable!
With this beta, tensor models (for Gaia, as the only tensor model that starts processing) finally work on RTX 3000 cards.
Happy New Year!
Happy new year all!
Great work, as all the others users, i confirm the increase in speed for the model that works (A4500).
Can’t wait to test others models 
Thank you for your great work!:
@suraj @nipun.nath @gregory.maddra
Warning! Harsh but necessary criticism ahead!
This is not specific to this beta but is relevant to all 3.x releases.
There is still major inconsistency throughout the program code as far as the format of the path strings. It is messy, ugly, sloppy and needs to be gone over and standardized.
Some examples, pulled from a posted log file (scroll to the right to and look at the path entries):
2023-01-07 10-51-39 Thread: 16784 Info OUT: 2 2023-01-07 10-51-39 Thread: 6840 Info putting fetched file into: "C:\\ProgramData\\Topaz Labs LLC\\Topaz Video AI BETA\\models\\"
2023-01-07 10-51-39 Thread: 16784 Info OUT: 2 2023-01-07 10-51-39 Thread: 6840 Info Created temporary file at "C:/ProgramData/Topaz Labs LLC/Topaz Video AI BETA/models//prap-v3-fp32-ov.tz" with 4602743 bytes
2023-01-07 10-51-40 Thread: 6840 Info Loading time for model file: "C:\\ProgramData\\Topaz Labs LLC\\Topaz Video AI BETA\\models\\/prap-v3-fp32-ov.tz" is 105095
Line 1 - double backslashes \\
Line 2 - single AND double forward slashes / + //
Line 3 - double backslashes AND a single forward slash \\/
To add to that, the registry entries are STILL written in a similar inconsistent state, after a clean installation, something I reported MONTHS ago:
For the love of this software, please get the VAI code into a consistent state! The current state is reflected in how bad the log files are!
Thank you!
Another Question: Need to know before I install.
Can the Beta co-exist with the latest ‘official’ release, currently TVAI 3.0.11?
Or do we need to do anything specific to keep them from getting in conflict with each other?
On Windows at least, the beta and official release do not interfere with each other.
This is because the beta installs to a separate folder C:\Program Files\Topaz Labs LLC\Topaz Video AI BETA and model data for the beta is saved to a separate folder C:\ProgramData\Topaz Labs LLC\Topaz Video AI BETA
I keep 2.6.4, latest beta and the latest v3 release on all of my machines. Up until this latest beta I’ve never had a problem just installing even without uninstalling for a very long time. This time I went and did a complete clean and purge of all installations and the registry as soon as Proteus didn’t work. That was a waste of time as the problem still exists on my 30 series Nvidia cards after reinstalling. I also use the install defaults to make sure I am using the same directory structure as Topaz. I don’t like deviating from the recommended when the software could be unstable.
Same here too. It hangs on “model loaded” and does nothing.
Out of curiosity, are you running the latest Nvidia card drivers? The newest ones (for RTX 3049 and 4909) was released on the 5th.
Delete all of your .tz profile sets so fresh ones will downlpad
Just doing a clean reinstall of the .tz files and refreshing/cleaning your video drivers may help.
Find the model folder and delete all .tz files. (I usually just clean out the whole thing.) - I’m assuming that this beta isn’t sharing the folder used for other betas and released versions.
Thanks. Where do i find this folder? (Windows)
The path should be in Preferences. (I think.)
