Offline Model Downloads

[4EvrYng (Alexander)]

@suraj

Here are the scripts for anyone wanting to download all models for offline usage. To use them open the terminal from inside the app and run the script. Feel free to modify the scripts depending on the models and input resolutions you need. Feel free to ask any questions below.

Hi Suraj, whenever I try to run your script ffmpeg crashes several times in turn crashing Windows. Error messages are below. Please advise what is going on.

---

Description
Faulting Application Path: C:\Program Files\Topaz Labs LLC\Topaz Video AI\ffmpeg.exe

Problem signature
Problem Event Name: BEX64
Application Name: ffmpeg.exe
Application Version: 0.0.0.0
Application Timestamp: 64469d9b
Fault Module Name: ucrtbase.dll
Fault Module Version: 10.0.19041.789
Fault Module Timestamp: 2bd748bf
Exception Offset: 000000000007286e
Exception Code: c0000409
Exception Data: 0000000000000007
OS Version: 10.0.19045.2.0.0.256.48
Locale ID: 1033
Additional Information 1: b6c5
Additional Information 2: b6c57350c7c460c42e0fab80ca95b2fa
Additional Information 3: ef98
Additional Information 4: ef98f3aafc1389b3e77d5632daf3e24b

Extra information about the problem
Bucket ID: 1fe1f87746768a256c97050fa74716c5 (2060120919334393541)

---

Problem Event Name:	LiveKernelEvent
Code:	141
Parameter 1:	ffffac8e3a968050
Parameter 2:	fffff80363960e60
Parameter 3:	0
Parameter 4:	3228
OS version:	10_0_19045
Service Pack:	0_0
Product:	256_1
OS Version:	10.0.19045.2.0.0.256.48
Locale ID:	1033

---

[badaconner]

Script doesnt work for me. Gives me ffmpeg error.

[Its_a_me_Mario (raj.b)]

It only worked when I edited it to specifically point at the FFMPEG binary in the TVAI folder. However, it takes a long time to run and I still only ended up with ~11GB in my models folder from a total of 302 files, not the >100GB that everyone else seems to be getting. I don’t know why that is. I’m running a laptop with a Ryzen 7 6800U CPU and Radeon 680M GPU. When I try to preview different configurations in TVAI, it inevitably wants to download the model despite making sure it’s configured to look at the same models folder I’m downloading to.

[barunendra.bachar]

This is very frustrating!!! Model download is taking very long time…

Please provide a separate link for downloading the Models at a time…

Thanks

[badaconner]

Really annoying to have to pay a lot of money for this software and not get the main models downloaded. Guess the company does not care. Definitely gonna try to get a refund

[tony.topazlabs]

Hello,

This script does download all of the necessary model files for every filter except stabilization. Is there a specific issue you’re seeing when using the script?

[stanislas]

Can anyone get this script to work in macos? If I run this from /Applications/Topaz Video AI.app/Contents/MacOS I get an output similar to what others have had and nothing gets downloaded. Any help would be appreciated.
Running Ventura 13.2.1 on an M1 Mac mini, Topaz Video AI 3.2.5

[badaconner]

Yea I am having ffmpeg errors like the others on this thread.

[olrik1953]

I tested the script for macOS and it doesn’t work. Why provide a script that obviously does not work and moreover without any detailed instructions. I am very disappointed with this version 3.2.7 and for the moment it is absolutely not worth buying until the version is at version 4. Version 2.6.4 remains for the moment better designed.

[chelsea.johnson]

I’m a Mac user and was running into the “Downloading Models” error. It seemed that every time I tried to generate a preview it would hang on that error and fail.

I wrote a PHP script that searches through the Topaz Video AI log files, regexes the missing models, then downloads them into the models folder. Any time I get that error, I just run the script and it pulls in the missing models. It’s been useful to me, so I figured I’d share in case it was useful to someone else:

https://gist.github.com/matthewjohns0n/cb525906e76ed6d5b7f09fc1e198e937

To run, download the file, cd into the directory the script is located, and execute php getMissingModels.php from the terminal.

[tony.topazlabs]

Hello everyone,

The Models Download Script has been updated to include the Iris model and to correct a name mismatch error when downloading Artemis Anti-Aliasing model files.

[644200]

[swscaler @ 0000016082A12040] [swscaler @ 0000016082B79640] No accelerated colorspace conversion found from yuv420p to rgb48le.
[Parsed_tvai_fi_0 @ 0000016080145B80] Model not found: chf-3
[Parsed_tvai_fi_0 @ 0000016080145B80] Failed to configure output pad on Parsed_tvai_fi_0
Error reinitializing filters!
Failed to inject frame into filter network: Invalid argument
Error while processing the decoded data for stream #0:0
Conversion failed!
Input #0, lavfi, from ‘color=c=red@0.8:duration=0.01:rate=100:size=1280x720’:
Duration: N/A, start: 0.000000, bitrate: N/A
Stream #0:0: Video: wrapped_avframe, yuv420p, 1280x720 [SAR 1:1 DAR 16:9], 100 fps, 100 tbr, 100 tbn
Stream mapping:
Stream #0:0 → #0:0 (wrapped_avframe (native) → wrapped_avframe (native))
Press [q] to stop, [?] for help
[tvai_fi @ 000001B4048F0E80] Unable to parse option value “0” as video rate

[joebnb]

this scripts not working in 3.2.2 @macos

[ida.topazlabs]

Hey there, did you launch from the app itself or directly from the terminal? You will need to launch it from the app directly.

[joebnb]

yes,i luanch terminal from app menu => process => open command prompt

[tony.topazlabs]

Hello,

From what I can see in the screenshot above, the script is running as expected. Could you post the last line before the script exits to see what could be causing an error? Thanks

[george.nagle]

Two questions:

1. Is there an update needed for 3.3.9?
2. Any chance of a step-by-step guide with screen shots or even a fast video on how to do this on a mac please?

[simon1234]

This is really appalling. I am on Topaz Video AI v3.2.9 and can’t upgrade.

The software itself is completely failing to download ANY models - tested on multiple connections and on 3 machines now. Presumably some bug has been addressed but now I can’t get that fix because I can’t afford to update right now on our charity’s basically non-existent budget.

This means that because all of the models are out of date, I am literally unable to use the software now because every attempt at a render or preview tries to download a new model. Now that there is no model mass download tool (what on earth are you playing at removing/not providing this functionality, given how often you force people to download models!), I can’t do anything about that in the software.

I came looking for help here and found this script. My models are all stored in user/AppData/Roaming/Topaz… so firstly, this script doesn’t check that and hence isn’t working. I attempted to work around this but as I’m not familiar with whatever you are using ffmpeg for (some download command I presume in there, akin to wget or curl perhaps?) I can’t see how to make it look at the AppData path… or at least I think I can’t - when I did get it up and running (possibly) it just sits there for 30+ minutes (I ended up terminating it in the end) with this output:

Input #0, lavfi, from 'color=c=red@0.8:duration=0.01:rate=100:size=480x360':
  Duration: N/A, start: 0.000000, bitrate: N/A
  Stream #0:0: Video: wrapped_avframe, yuv420p, 480x360 [SAR 1:1 DAR 4:3], 100 fps, 100 tbr, 100 tbn
Stream mapping:
  Stream #0:0 -> #0:0 (wrapped_avframe (native) -> wrapped_avframe (native))
Press [q] to stop, [?] for help
[swscaler @ 00000197A2321880] [swscaler @ 00000197A23AE480] No accelerated colorspace conversion found from yuv420p to rgb48le.
[swscaler @ 00000197A2321880] [swscaler @ 00000197A23BC540] No accelerated colorspace conversion found from yuv420p to rgb48le.
[swscaler @ 00000197A2321880] [swscaler @ 00000197A23CA600] No accelerated colorspace conversion found from yuv420p to rgb48le.
[swscaler @ 00000197A2321880] [swscaler @ 00000197A23D86C0] No accelerated colorspace conversion found from yuv420p to rgb48le.
[swscaler @ 00000197A2321880] [swscaler @ 00000197A23E6740] No accelerated colorspace conversion found from yuv420p to rgb48le.
[swscaler @ 00000197A2321880] [swscaler @ 00000197A23F4800] No accelerated colorspace conversion found from yuv420p to rgb48le.
[swscaler @ 00000197A2321880] [swscaler @ 00000197A24028C0] No accelerated colorspace conversion found from yuv420p to rgb48le.
[swscaler @ 00000197A2321880] [swscaler @ 00000197A2410980] No accelerated colorspace conversion found from yuv420p to rgb48le.
[swscaler @ 00000197A2321880] [swscaler @ 00000197A241EA00] No accelerated colorspace conversion found from yuv420p to rgb48le.

But zero sign it is doing anything. It won’t even allow Control+C quit, I have to force quit via a task manager!

This is not cheap software and we shelled out for the complete suite. We knew it would expire and likely we couldn’t afford a further upgrade but we at least expected it to stay functional at the point it got to with updates on our licence!

Even when it was downloading models, the way that you service this software is horrendous. Virtually every time I’ve come to use it (often in a hurry) I’ve had to sit through huge model downloads or software updates, which on my connection can often take an hour or more - the software is completely unreliable in a busy workflow, absolutely shocking!

To make matters worse - it seems (presumably because my subscription ran out) that every avenue for contacting customer services is set to divert me to an FAQ with no option to actually contact anyone!

[finnius1]

I would love to be able to just download the models without using a script. I remember 2 years ago when that option was part of the program. It would even show you older versions that you could grab as well. Why not just re-implement that back into the program? There are many many times when I prefer to be offline and it is so much easier just to have the full model package on my system. Having to keep online while I might be spending a day or more working thru a queue of tasks seems silly. I am sure others would agree with me on this point.