Topaz Video AI v3.1.11

Topaz Video AI Releases
22

I understand, but it would also be nice to update the other base models. Because they are no longer updated since 2.6.4 like Artemis, gaia etc. Why not they create artemis 14 for example and that it would be better than 13 :slight_smile:

2 Likes
23

The numbers are just to organize and have a marker for the changes in the software. So 3.1.11 after 3.1.10 means that the version 3.1 just got a minor and “cosmetic”/urgent change. And we could assume that 3.2 could have a big change, like add/remove features and major bug fixes, etc.

3 Likes
24

I have 1TB of my SSD

25

Always appreciated with a new fresh update!
Regarding Tiff export, a small box where you can click in export wav of the same trim would be appreciated.

26

In your opinion, is it better of quality to export to video or to 16-bit png or 8-bit jpeg? Why if I export in image sequence, png 16 bits does it take longer to encode? In video it goes faster. Explain me ? thanks

27

Sorry, what I want to mean was, how much free space do you have in the OS drive?
Because when you are upscaling or downscaling the software generate several large files in the AppData\Local\Temp. I didn’t finished my tests, but when I was downscaling 8k to 1080p it started to spam 48mb files there and it was too much to my HD to deal, then the system crashed. If the process you are using does something similar, this behaviour could also overheat your SSD and that could lead to read/write errors.

28

Think of it more like they have a huge list of things they need to develop. They brake those down into small sections that can be worked on in a short amount of time and tested. The work will take as long as it takes no matter what. The weekly releases are more just a bonus of: what’s done gets released, and can therefore make in into the cycle of being used by others and praised or condemned.

29

From what I understand they are putting most of their resources into improving the preview functionality right now.

Of course I’d like to see v13 (or would it be v14) Artemis, v4 Proteus and v6 Gaia models eventually too :grin:

30

Damned if they do and damned if they don’t – no matter what they put in a weekly release half of you will be complaining about it.

Thanks for the update, small incremental fixes are better than radio silence.

8 Likes
31

Personally, if you export to a 2TB SSD in image sequence or video format, the speed remains the same. Likewise, if you recompile the image sequence from a video software (if the image sequence is on an SSD) it will be as fast as a video sequence. On the contrary, on an HDD, it will be slower than a video file because the difference in read/write speed is quite big compared to the SSD.
To keep a very good image quality, I recommend the move format for the video file and the png for the image sequence. The Tiff format is much too big. And I didn’t see any difference with the quality in png. Especially since Tiff will slow down the compilation in another video software since it is heavier to process.

32

Frequent releases squeeze normal development time, and also make development more inclined to simple work, like the above-mentioned “only GUI update”, because it is easy to do. And even though it’s simple, Topaz doesn’t add a single quality control parameter.

If every update of Video AI is like the last two, it is better to reduce the frequency to half a month or one month, it is too superficial. Reducing the frequency can also accommodate some people’s comments that “frequent updates look unprofessional”.

1 Like
33

So far I still have the issue where CPU is selected to process regardless of which gpu I selected. Annoying issue that makes me not use your application.

34

So in other words, the software is half broke or half fixed?

35

That’s mine thoughts too.

36

This is indeed the simple yet so much required function.

37

Upscaling to 4K appears to be broken. When upscaling a 1440*1080 video, the output adds black bars on the top and sides and puts the video in the center. Does anyone have a link to the 3.1.10 version so I can test the same file?

38

It’s the video memory problem. To be clear - lack of video memory.
Mine GTX 1050ti with 4gb VRAM simply cannot work with the large videos - upscaling to 4K or Chronos model with Fullhd videos.
There is no solution to this except revert to 2.6.4 version.

39

My GTX 1060 3GB could do it if I set the max memory to 10. The time it took though, that’s another story.

40

You set your video memory usage to 10%?

41

If I recall correctly, TVAI uses ONNX to deploy their AIs to the end user. Support for AMD AI accelerators will come when ONNX gets an update to support them, and TVAI updates to that version of ONNX and enables the feature.

Note: I might be wrong about this.

1 Like