Offline Model Downloads

4.0.1
Following all instructions
I’m getting this issue for every model in the script:

[AVFilterGraph @ 000001df82c7b780] No such filter: ‘tvai_up’
Error reinitializing filters!
Failed to inject frame into filter network: Filter not found
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=720x480’:
Duration: N/A, start: 0.000000, bitrate: N/A
Stream #0:0: Video: wrapped_avframe, yuv420p, 720x480 [SAR 1:1 DAR 3:2], 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

2 Likes

i am very non technical when it comes to executing this install. do you have a video tutorial or written instructions on how to execute this bat file in topaz ai video terminal. any information on how to do this would be greatly appreciated

2 Likes

OK- so why not give us an option to download all of the models in the application so that they are loaded properly and we don’t have to mess with scripts?? The models for the Photo AI has the models download for each update, so I would think this is possible, right? I get the models for the video are MUCH bigger, but honestly, warning people about that before they just download all of them is enough, I would think, compared to DL everything individually or messing about with a script.

I do recognize, maybe I am missing something. Maybe there is a super valid reason for this, but would like to know what that is.

3 Likes

Here, I recorded a little video showing you how to do it: https://youtu.be/AkCj54kGIWw

7 Likes

Cheers for that effort. (sadly mac here)

1 Like

Aw, dang :frowning:

Updated for Video AI 4.0.0

But the script v4 still says:

#!/bin/bash
# For version 3.2.2

and nothing about v4 at the top of the script.

The issue with the script not working seems to be a permissions issue. TVAI does not give itself permissions (even if you run as administrator - obviously I’m on a Windows machine here) to install to the default models folder so every single attempt by the program to download a relevant model fails. Either you need to run the script as described above and relocate the models folder to a totally different location that doesn’t need the permissions, or (if you want the program to work correctly without needing to run the script), go to folder properties, security tab, and give all permissions to trusted installers and to the users.

You can actually use DLVideo’s tutorial – the only difference is to go into the sh file via TextEdit and replace both instances of “ffmpeg” with the full path:

“/Applications/Topaz Video AI.app/Contents/MacOS/ffmpeg”
(it may be “./Applications/Topaz Video AI.app/Contents/MacOS/ffmpeg” – am not remotely skilled at this stuff).

Run the sh file and it’ll download the models. Head to the folder to ensure it’s populating, as the terminal window will look a mess, what with all the “No accelerated colorspace conversion found” notices.

For anyone having trouble with macOS:

1- Download the macOS script zipped file from the first post on this page. Extract the downloadModels-v4-1.sh script file and copy it to

/Applications/Topaz Video AI.app/Contents/MacOS

It should be sitting right next to the FFMPEG file (remember to right-click “Topaz Video AI.app” and selecting “Show Package Contents” to be able to get to that folder)

2- Give “Full Disk Access” permissions to both “Topaz Video AI.app” and “Terminal.app” in

macOS → System SettingsPrivacy & SecurityFull Disk Access.

After doing that, make sure you completely close both apps (or even better: Restart your computer!)

3- If you want to download models to a different directory (or external hard drive), open Topaz Video AI and go to

PreferencesDirectoriesModel Folder

browse to select your new Model Folder location, hit “Save”, and then make sure you completely close the app and re-open it.

4- Open Topaz Video AI app and go to

ProcessOpen Command Prompt

A Terminal window will open. Then type the following:

sh (and hit the space bar to add a gap)

Then drag and drop the “downloadModels-v4-1.sh” script file (from the /Applications/Topaz Video AI.app/Contents/MacOS folder to which you copied it after downloading) to the Terminal app

Then go back to Terminal and hit the Enter key to execute. This is how it should look before pressing Enter (in case you want to copy and paste directly):

sh /Applications/Topaz\ Video\ AI.app/Contents/MacOS/downloadModels-v4-1.sh

Download is slow, give it time. You’ll know it’s done when Terminal shows you the default “%” sign again.

I hope this helps! There’s no need to modify the script!

3 Likes

Please use the model manager inside the app.
The 4.0.7 version of the app now has model manager to download the models.

3 Likes

Doesn’t work for me, my connection speeds are fine, I’ve had to download models manually through the log files since April 2023.
Desktop Screenshot 2023.12.13 - 18.53.29.95

1 Like

No problem here with the model manager (in 4.0.8.0.a that is).

The software just hates my PC then. lol

Same issue here:
image

Maybe a Windows / Mac difference (I’m on MacOS)?

Or related to specific GPUs / specific models? I only tried those that I do normally use: Iris, Proteus.

So: maybe try, if all models fail to download or just specific ones. Do you have a Firewall / VPN / Antivirus installed ?

@TheRustyRebel @notes4me Make sure the GUI has access to the internet, seems like firewall is blocking connection. Make sure the app has https access to the following hosts: veai-models.topazlabs.com and models.topazlabs.com

2 Likes

Is this what you mean or something else? In April 2023 models refused to download within the app and had to get them manually from the log files, since the 4.0 release it now downloads models if like it used to but takes 10x longer than it would if done manually. I thought the model manager would do the same job but appears to fail, if it were a connection/firewall issue then surely it wouldn’t download models while I am trying to use the software before model manager was released?

It uses up to 40-Mbps when I start the download in model manager but fails after 10-20 seconds later, doesn’t make a difference which model or how many I select either.

can you send me the application logs