If they are privy to information thst others do not have then it would make sense to keep them separate.
There has been changed so much in therms of A.I. Developement in the last year.
Because people start to understand whats happening and how a neural network does handle its data.
1 Like
Despite its problems, TVAI 3.X has been the best thing to happen to my workflow since I discovered the product over a year ago.
3 Likes
As it stands, TVAI v3.0.5 does not compare to VEIA v2.6.4. TVAIās performance and results are a lot worse, upscaling isnāt as sharp, interpolation has more artifacts and makes the output blurry for some reason, and the time it takes to process everything takes significantly longer. Even the compare view is gone, like why?
3 Likes
Where? I looked and canāt find. Thanks
I agree with you, version 2 seemed sharper to me, but i will say this, i did some experimentation in version 3 and i noticed that proteus was blurrier than artemis, which is crazy because in version 2 proteus was sharper. so in version 3 i would suggest using artemis low quality for getting it sharper until they fix proteus to match how it worked in version 2
1 Like
Here is a sample showing how sharper artermis low quality looks compared to the original dvd quality. Artermis looks good on old dvd quality in 3.0 from what i can tell. hope this helps others. Video AI Team, please fix the proteus model in 3.0 because in 2.0 proteus was the sharpest and clearest model but it doesnt really have that clarity in 3.0. thanks for all your hard work, i know its not easy to work on this for months. We all want it to be the best and that is why we provide the feedback. thanks again
1 Like
I disagree with that, but saying that without even mentioning what model your using is pointless. GHQ is still fine, Proteus actually looks better and I find myself using AHQ now which I did not in the past. DDV is great when used on interlaced material. I got over trying to upscale garbage quite some time ago so I have no idea how that goes.
When they first released v2.x with chronos I tried it and chronos took forever and went back to using flowframes. Now that Chronos can be ganged in with an upscale, and itās actually way faster considering, I donāt use flowframes at all anymore, unless of course the video doesnāt need upscaling. I have not tried V3 on my AMD 6900xt since the first release version but it was way too slow on AMD, itās much better on Nvidia. V2 was fine on AMD
After getting used to V3, I can say itās pretty good. There are a few interface and workflow tweaks that need to be addressed but I still have V2 installed on the main video box and I never use it anymore.
Proteus in TVAI v3.0.5 is worse than VEIA v2.6.4. Itās not as sharp even with the same settings and takes about 100% longer to render (1h 15m from 1080p 60fps to 4K before, 3h now). Proteus has always been the best upscaling method, many people donāt know how to use it and wrongly claim Artemis or any other model is better for generic upscaling. Comparing the two (custom settings for Proteus and Artemis) in VEAI, Artemis looked like it didnāt do anything before, which it still does in TVAI, comparatively. Chronos Fast has more interpolation artifacts than the earlier version and blurs the output video which it didnāt before and like everything else, takes longer (45m for 1080p 24fps to 60fps before, 1h 30m+ now). Iāve used VEIA consistently for the past year and a half for interpolation and upscaling and I can tell without comparing (I have BTW) that the results and performance of the latest version of TVAI is poor. Iām not planning on switching my workflow until I can get equal or better results. I have an Nvidia GPU. Donāt get me wrong, I love Topaz and their products, but TVAI needs to be overhauled if it wants to compete with its predecessor.
2 Likes
This is only half true.
Artemis models in around version 2 and newer are worse - any Artemis model newer than V8, which was most of them through much of version 2ās release - is worse than Proteus in most circumstances. However, Artemis V8 and earlier, meaning you have to use older versions of Topaz, were significantly better in detail enhancement than Proteus can get right up to latest 3.0.5 release of Topaz.
I also have sunk an insane amount of hours into Proteus since its inception, and I can rarely match the output of the older Artemis, and when it gets close, it gets more artificial as it doesnāt enhance fine detail well. I have been trying, and failing, for months and hundreds of tests now to try and get any of the Proteus versions to outmatch the older models of Artemis without success.
The sad state of it is that due to Artemis being broken around version 2, Proteus is the best upscaling option current in the program, so the statement is correct for most people - its just that its become that way because the older Artemis models donāt work anymore and the quality reduction seems to have been accepted to prioritise user manual sliders for adjustments.
I suspect this is largely a side effect of Topaz trying to move away from models that need a lot of adjustments prior to importing the media (what you fed in makes a large change to the output) to a model solution that tries to make it a āone stop shopā for a general user.
I donāt disagree with that end goal, but it did sacrifice quality in the models to achieve it presently. I am just patiently waiting for improvements to the Proteus models to catch up to where Artemis was originally. At that point, then it will be the best model for upscaling.
I was using the 2.x version almost all day and for up to 18 hours nonstop for one file after another. Since version 3.x my work stopped. After months hoping vor a 3.x version that at least matches the usability, speed and output quality of version 2.x I feel totally screwed up not being able to continue my work either with the old version because any further support is abandoned nor with the new version that lacks fundamental. Or to say in different words: we are still waiting for the manufacturer to find the perfect color tone of pink with a note of magenta in between instead of getting the damn porsche to run, which it should normally be made for ā¦ 
1 Like
It would be interesting to hear about sales.
Because V3 looks like some GUI for some command line tool made by geeks for geeks. It looks like some student project and works like some student project. And it is not good for salesā¦ To sale something you need product with first look impression, style and some magic inside. V2 had magic and style and V3 absolutely not.
Seriously? Bilinear frame resizing in previews? And how can we call it āpreviewsā? If you canāt make normally working preview, just remove it until it will be completely fixed. When you see glitches in first minute of testing, you never make a decision to buy something. This is the basics.
So, it would be sad to lose such promising Topaz team with unique products because of financial failā¦
1 Like
