Topaz Video AI v3.1.2

Topaz Video AI Releases
1

Hello Everyone!

A new release of Topaz Video AI is now available.

v3.1.2
Released January 31st, 2023.

Download: Windows, Mac.

Changes from v3.1.1:

  • Improved Auto-Crop or Full-Frame Stabilization for long videos and videos with shot changes
  • Significant performance improvement for GTX 16xx series GPUs compared to 3.x.x
  • Support for single channel image sequences
  • Fixes Apollo frames being less than expected
  • Fixes Apollo getting stuck at slow-mo factor over 10x
  • Default for video export setting is now “With Preview” that fixes exported video not opening in other applications
  • Misc. bug fixes and improvements

Known Issues:

  • If all models fail, please try deleting the .tz files in your model folder. The location of this folder may be found in the preferences menu

Submit files here: Dropbox

Please take a look at the Video Roadmap Update for more context on what we’re focusing on right now. Thanks in advance for your feedback!

12 Likes
2
3

Thank you for updating every week. Where do I ask questions about settings?

4

Ill need to download and test this version but I noticed in 3.1.1 and its still present in this version, that anytime I add noise or grain it only applies to the first frame then it turns very blotchy with each subsequent frame thereafter. Its as if it is being applied before the model is rendered, and the model is trying to denoise it.

Did anyone else experience this?

1 Like
5

Can someone with a RTX 4080 confirm what encoding speeds they are getting with a 4K video with Proteus?

With v3.1.1 I am only getting 1.5fps

6

Sadly - still the same red vertical lines appear when I export to ProRes - so regrettably sticking to 2.6.4 which works. It’s a shame as v3 is starting to work really nicely for me with the cropping and the non-strobing de-interlacing.

1 Like
7

In the Topaz Video AI Discussion.

8

Not sure where to post this, but I am getting a “flicker” on my screen when using this version of the app.

it only happens when i have the app on my screen, and only in the app itself. Seems to be tied to the mouse movement because if i leave the mouse still, it doesn’t happen. Doesn’t matter if I am actively encoding or not–still happens.

other than that, this version is running great so far. getting solid 13 fps

1 Like
9

From what res are you trying to going from?

10

Do you have freesync/g-sync enabled for windowed applications? Some times they can interfere with things like this.

Although, since you claim it happens in this version of the app, and not other versions. It would suggest this change is related to changes made by the developer and maybe they should look into fixing it.

1 Like
11

Why Do I have to DLd models after each update? I have limited bandwidth and using my plans data.
it’s getting really old especially with an update every day or so. need a way around this.
Mike Cossey

1 Like
12

Because they have changed and been updated.

13

Not upscaling, trying to export a 4K using Proteus for purely to remove noise + sharpen.

14

RTX 3070 around 2.2fps - 2m23s

Proteus, no upscale, Relative to auto all at zero except sharpen at -100

Tested on this stock footage:

1 Like
15

It depends on your source video, target video and model.

16

YESSSSS!!! FINALLY! " * Significant performance improvement for GTX 16xx series GPUs compared to 3.x.x" I’d say so!!! From being stuck with 3.0.12 because it was 7 times faster than the incredibly slow pre-3.1.2. This is awesome now I’m upscaling some DVD TV episodes that were taking me 2 hrs and 15 mins with 3.0.12 and now upscaling with 3.1.2 the same thing at 1 hr 5 mins. Lovin’ it. I just hope my grain settings are right since the options changed from 3.0.12.

17

This is pretty awesome, I agree! My GTX 1060 3BG was getting 0.02x playback speed and is now getting 0.14x! (No idea what that translates to in FPS)

18

is there a section just for suggestions, a same as source would be handy for bitrate, just sayin thanks

19

I’m running out of memory on jobs that have successfully been completed with previous versions of VAI. M1 MacBook Pro with 64 GB memory.

20

In v3.1.1 I was getting out of memory when running a batch job of say 17 or 10 files. Those that failed when rerun exported ok.
TopazVAI was set to to perform 4 processes. I will try less in future.
I have reported a bug to Topaz on this issue.
I have a MacPro with 96Gb which only got to 50Gb’ish looking at Activity Monitor.