To my great regret, Proteus ‘dithering’ issue has returned. 
Look at this:
Or this:
And here it’s downright horrific!
Still using RTX 3080 Ti. And deeply saddened, once again. Unusable in this state.
1 Like
I will agree with you, that the amount of bugs is not only staggering, but, worse, that we’re basically still experiencing a regression from v2.6.4 – which is decidedly not good. For every door they open, another one closes again.
5 Likes
Sadly, that seems to be the case. I don’t mind waiting for the bugs to be fixed and performance for the 2.6.4 restored and or surpassed. And I like the road map Topaz published. But their management of those expectations must be tempered with reality of the current buggy software. They should keep it in beta, let people who want to test it for free and when its all ready for prime time, release it with the license they have now. If it takes them another six months or a year it would be realistic time frame. I would not mind waiting, but I’m not going to purchase a beta under premium license. Thank god for the forums here so I don’t get a nasty surprise as some have. Kudos to the community over here for posting about their experiences. As for Topaz, they need to do better. Bugs like we have now are forgivable… in beta, but not in full software release.
4 Likes
Unfortunately I am getting “Model Loaded” and sitting indefinitely. Running as admin, uninstalling/reinstalling does not help.
I changed the AI model away from “Proteus - Fine-Tune/Enhance” and then the program worked fine.
It seems there’s a bug in version 3.1 with Proteus that prevents it from working at all? Anyone else run into this? Any solutions?
1 Like
Hello,
It’s great, but so far professionals, participants in the Berlin Film Festival, do not use Topaz.
Probably because Topaz is unlikely to be able to make such a fragment Police (1916) Charlie Chaplin - YouTube
With respect for your activities,
Eugene
1 Like
For those looking to improve speed, a little hint.
Use a seperate drive for your source and output.
- Read and write to same drive 7.0fps Time: 01.07
- Read one drive write to another drive 8.1fps Time: 00.57
2 Likes
same here. all modes work fine. except proteus that runs stuck at 83 % of preview and won’t start processing either.
I also think the decision to end the extended update supports to end where it did, with the issues that everyone was having due to the new version to be very disappointing. The whole reason this was done was due to the condition that 3.0.X was in and has been in for so long and finally it appears you’ve fixed and improved things (Not to mention features being brought back in line with the old version 2.X.X in this version) to a point that it appears to be ready to be an end point, but you decided that it was going to be a new starting point and now the end point was still left in a rough way.
All of this aside, the program should not have a one click auto update button that takes it to a version the license doesn’t support. This need to be incorporated into the program to do checks for eligible versions based on the account logged in. I have never used an automatic update in a program that has left me with a version that was unusable before.
6 Likes
When a similar event occurred in the past, it was a color space conversion error resulting from the conversion between Rec. 601 and Rec. 709.
By examining the resolution and color space of the source video and the resolution and color space of the output video, we may be able to guess the cause.
Let us know what the reply is. I went back to 3.0.7 and had no issues. Reloaded the new version fresh- and had a new error! Now it wouldn’t even let me import the video file!
I am impressed by the themis model which deblurs the videos really well. and in addition it denoises without losing too much detail. for a long time I was producing in SD (original size) and I was editing with vegas pro with an HD project, but the fact of going back to SD with vegas gives me the impression that I find more details strangely… on the other hand I uses a gtx 1660, 16gb of ram and a ryzen 7 2700 but it’s much slower than 2.6.4… 4m55 for 2 seconds preview in SD with the themis model… (1 to 2 spf, and even with the artemis models it is long). Lowering the ram in the preferences doesn’t change anything either.
How do you mark I/O in the timeline if you only want to process part of the video?
What upsets me the most, is it came without warning. Topaz absolutely knew 3.10 would end use of the software for many users, with expired licenses, but still they said nothing. No, they just let folks go through the hassle of downloading and installing 3.10, and then the big surprise. The surprise should have come long before the download part. That’s pretty poor customer service if you ask me.
6 Likes
but it was said when have to buy the license that it was valid for one year. so yes we could still use the beta, but it’s a major update so it’s logical that it expires since the license was valid for one year. We’ve always been able to get the latest betas without a watermark, that’s not bad. So me too if my license has expired but getting full won’t change anything, that’s how it is… we can still use it even with a watermark, it’s still better than nothing 
I’ll say two things about the expired licenses:
I knew full well before I clicked the buy button that I was only buying the upgrades or updates for a year. I was just glad it was not a subscription service, and I was happy with the prospect of still being able to use what I got without updates after my year was up.
The only other software I own that has a similar pay to update model, won’t even let me get to the update without renewing my purchase. That is more what I expected with TVAI, but not the case. May Topaz learn the value of such communication and implement it.
4 Likes
Man, what is this?!
Not counting the slight frame-disparity between source and preview, where is all this ludicrous dithering coming from? And that’s with +20 noise reduction (which is already too much under regular circumstances). ‘Recover detail’, you think? Nope. Does this even at 0 recovery of detail.
Getting despondent again. Output can’t be worse than the source, guys. TVAI should adhere to the Hippocratic Oath: “Above all, do no harm!”
2 Likes
Wait, wut?! Yeah, no.
Out of curiosity what does it look like using Auto?
Far worse, even. Earlier, I pretty much used the ‘estimate’ values, with added noise reduction (and a pinch of extra sharpening) to reduce some of the heavy artifacting you see here:
EDIT: Since I read here that people love to look at a woman’s lips, rather than staring at a wall, LOL, even with the best possible settings I could come up with, there’s still undue dithering taking place:
Can I give you a suggestion for values based on what your showing with the “lips”. I always found with noisy input files you need to lower the sharpen not raise it. So here are the values I would try.
4 Likes
