I think naming it version 3 was also a marketing stunt, which gives the impression it’s tried and tested software, but in reality, it’s not. Like you, they should have continued to support model improvements for 2.XX, while working on version “1” of the new software product. I’d still be using version 2, had they continued to make model improvements, but after giving up on SD video upscaling in version 3, I totally removed the software from my computer, but I continue to use other Topaz programs.
4 Likes
Absolutely on the spot.
I installed the version 3.XX and what it came to mind was: well, they managed to ruin a perfectly functional software to something that isn’t comfortable for me to use.
I didn’t bother to renew my licence… I had it for one year but I decided to stick to the 2.XX that does a pretty good job without the hassle - not forgetting that for me and my system, the 3.XX runned slower than 2.XX.
They did the same with the new Photo Ai. I hate it. I’ll rather stick to Sharpen, Gigapixel and DeNoise stand alone apps than let Ai decide which enhancements my image needs.
7 Likes
An interesting thing about this build.
In the 3.1 beta, performance increased when TVAI was in the background, this is not the case in the stable version. But when I turn off the screen the performance goes up by ~~20%. When I turn on the screen I can see for moments that I have much more FPS. And this is not a bug. I fire upscalation and it says it will finish in 4h. I turn off the screen, leave the computer to do its thing and it finishes much earlier.
Why would you remove a programme that is now (at last, finally, took them long enough) surpassing the 2.6.4 version?
So the way I see it, I updated to 3.0.0 (probably 3.0.3) about 4 days before my year was up. It was trash and I was annoyed but then Topaz due to 3.0.X being next to useless let us folks that had come to an end get further updates.
That’s fair and nice of them. So yes the past couple of 3.0.X updates have shown improvement. Just a few days ago I was thinking “yes this may be worth paying another year for!”
Then 3.1 0 came out and I updated to that. BANG…no go, your license expired in November.
Ahh okay says I, would have been nice of you to advise that when going to install! No worries I’ll just go back to 3.0.12 or whatever.
Oh no…we are just limited to 2.6.4 and any updates after (even ones within my year) I can’t have.
Is this tactic going to get me to part with my cash in these increasingly hard times? No.
So essentially we were just another bunch of beta testers who have now been thrown off the program, I would have been happy for Topaz to have just advised us “temporary betas” that we would be cut off from 3.1.x line but could remain on the 3.0.x line. But no, thanks for the feedback guys, now duck off!
4 Likes
In light of recent … issues regarding licences and updates/upgraded and the discussions here about the quality of the TVAI and Topaz in general… I’ve just happily renewed my own licence.
Have a nice day.
I’ve already explained my reasons, and for now, that’s the way it is, but I’m glad it works for you.
1 Like
Reinstall 3.0.12 or whatever over the top of 3.1 and it should work just fine (as a beta!). After trying 3.1 in trial mode, I did that and 3.0.12 shows up as licensed. The nag popup can be dismissed for 6 months - perhaps by then it will have achieved release candidate status!
All that remains to irritate is a small “Update Available” box at top right of its menu bar.
It seems Motion Deblur (Themis) model always uses GPU 0 even if I set GPU 1 for AI processor in preferences.
They wont go on, says my licence has expired. I’m locked out going to 3.1.
I can instal 2.6.4 but if I go to 3 or any other (which fell within my years subs) it wont let me activate.
What a mess!
Edit, okay after uninstalling everything (2.6.4) and reinstalling 3.0.12 over the top we finally have got back to 3.0.12.
I stand corrected and chastend…for now!
1 Like
good find. did you try to limit the frequency of your monitor, maybe lower it from 120 fps to 60 or to 30 so the gpu has more time for processing?
1 Like
I’m also seeing significant performance degradation with Chronos Fast on an M1 Max MacBook, a video that took 27 mins to process now takes 45 minutes. But on my RTX3080 PC I got a significant speed up, 27 mins to 13 minutes.
Thanks for confirming. I’m fairly certain it’s because of this:
Chronos, Chronos Fast, and Full Frame Stabilization uses TensorRT models on NVIDIA machines
I think the change to TensorRT only benefits the RTX series of cards, anything else is now performing worse. Hopefully they can bring back the performance to non RTX systems.
1 Like
Windows 10
for couriosity upscale a hd 720p video to full hd 1080p takes the same time than upscale it to wqhd 1440p. 
1 Like
Hi - I’ve found that any upscale between 120% and 239 or 240% takes about the same time! There is then a big drop in speed above 240%. I’d be interested to know why that is.
Thanks.
Andy
I can answer that. TVAI has 3 different AI enhancement models, 1x, 2x, and 4x. Everything is enhanced through those 3.
1x works up to 120% scale, 2x works at 121% to 239%, and 4x works from 240% and up. 1x is the fastest, 4x is the slowest, and 2x is in between.
I also find that generally, 2x has the best quality.
4 Likes
i tried also with keeping the original size but the time was still almost the same.
It depends. Artemis at 2x may be about the same speed as Proteus at 1x, because Proteus is much slower.
This is why if I have a true 720p video that is not the best quality I will downscale(loseless) it first in something like Staxrip to 480p. I find I really didn’t lose much info because the 720p was poor to begin with. Then I upscale in TVAI to to 1080p. It’s much faster going to 1080p this way then 720p or 540p. Again though only if the original 720p video is of low quality like you can’t make out individual strands of hair on a persons head. Also I’m only using Proteus.
1 Like
