Topaz Video AI v3.0.6

Yes, you have already that option:
Split your file with thousands of frames into smaller segments and run as many segments in parallel as possible. That’s clumsy but feasible.
Automated scene detection by VAI, parallel processing of segmented scenes, and automated put together when all scenes are processed perhaps could be a solution. I support such an idea to be evaluated by Topaz devs.

So…reading between the lines…Topaz needs to employ some better developers/coders?

Maybe version 6 or 7 might do it, if they haven’t run it into the ground by then. :rofl:


you think you’re funny but you’re not.

it’s not about better developpers, it’s about bad timing, some bad company decisions (release of 3.0 done too early) and to achieve what you want, need much much more employees to have it available quick. topaz is a small company, and what you ask need a lot of money and people. and even with that, look at microsoft, they can have thousand of developpers, they release stuff with bugs and issues all the time.

want something worst ? take a look at the Nvidia Shield developpement and keep in mind there is Nvidia and Google behind it !

what you want is a turbo button like in Mad Max. maybe it will happen, but you’ll have to wait a bit. software developpement take times, and in case you didn’t noticed, this is the best software on the market to do upscaling. maybe it will change in the future, but for now, it’s still a fact.


I don’t think I’m funny. Just annoyed. Version 3.x.x already and it just gets slower and slower.

Maybe spend less time on added features no one has asked for and instead spend the next 6 months on performance improvements and bug fixes?

But maybe touching a nerve or two might get some results. But I’m just a consumer, what would I know?


Why does Topaz use its limited resources to add new features half-heartedly instead of focusing on improving existing features?
Users have a right to voice their dissatisfaction with that.


Is there any user out there who has experienced a performance increase compared to 2.6.4? I am really wondering and would be interested in their computer setup!


I agree with Martyprod, that the developers are not necessarily to blame here. The developers I ‘spoke’ to in email have all been super polite, helpful, and went out of their way to solve my issues. Their sales department, eh, that’s another story. :upside_down_face: VEAI 3.x was released too soon, too buggy. And as a fellow customer, I share the sentiment you are having about being mildly upset about some of the current, outstanding issues. Especially since those didn’t exist in the last 2.6.4 version.


It’s very inconsistent at best. So many things play into it: What source, what model, what output resolution.
One time, I ran a small test clip in the GUI and it said it processed it at 15fps, but when I copied the command and ran it in the CLI it said it was processing at 5fps.

For a sanity check, I just ran a 15 second clip on both versions using the Proteus Manual model. Same settings. 3.0.6 took 1:07 and 2.6.4 took 0:45. I tried about 3 runs on each and the times were similar each time.


For serious? That is quite a significant difference. And one I don’t quite understand. I mean, I assumed they just took their Proteus 3 code from 2.6.4, and ‘moved’ it to the 3.x versions. Guess not.

1 Like

Still a bug, perhaps an additional:

Previous versions would generate solid light pink screens with stabilize PLUS either option for rolling shutter/reduce jitter.

This new version caused the same issue without those settings. In this case adding 60 FPS interpolation to stabilization.

Second: I still am waiting for a fix for remote desktop to work again.

Rapid releases are great though, progress is being made. Hoping to get this ironed out so we can re-visit a Linux version.

What do I need to do to become a beta tester? More than happy to send in files/logs.


I perflectly understand the degree of non satisfaction from everybody, i’m as well a user like everybody else.
developpers are actually working on Performance. (main dev left a message on the beta test thread yesterday). They are (really) trying their best, Really !!!

does the software work right actually ? no.

people paid for it, and it’s damn expensive. it’s not the little software shareware at 30€.
The devs are aware of the situation and really does their best. i’m here since long enough to know that.
i can tell at 100% they are as much pis*ed as you and us about the situation. but they are not the ones who took the decision to release the software too early or in this state, as yes, what has been released is more closer to a beta version than anything else.

Now since 5 months… how many message like yours are we getting on this forum ?
a crazy amount of messages of this kind since 5 months.
after a while, maybe it’s time as users to ask ourselves a question. what this place here is for ?
we understand when it’s a brand new customer, he’s not aware of the situation but what about the others ? and some are there since 2 years or more !

the overally situation is toxic …the company, the developpement of the software, decisions taken, etc … we know that… but at the end of the day… it’s a lot of time lost for everybody to tell exactly the same thing over and over again instead of concentrate about reporting issues, and hoping it will help the devs to achieve the software we’re all expecting to have.


We do beta testing with TOPAZ products, well knowing that beta releases are not (yet) perfect. TOPAZ welcome our feedback but we are not responsible to manage what they intend to do. Nevertheless a roadmap of what is in sequence of development would be helpful for us users.

I’m now using TOPAZ products since about 2 years, and participate in this user forum. In summary I found that most of the complaints here result from exaggerated user expectations on AI to deliver magic.

TVAI is a learning process for the developer and for the user. However I also miss some guidance on how to tune hardware configuration, and model presets to improve processing speed and output quality. I’m aware that restructuring VAI from single frame in v2.x to 3.x multiple frame processing impact processing speed.

The variety of video codecs and compression formats evolved in the past 30 years is a huge challenge to train AI models to fit all. I hope (and expect) that TOPAZ maintain a statistic of user demands to fix the most requested issues first

1 Like

Well sometimes consumers getting a bit up in peoples faces does cause things to get better.

I hope to see some real improvements going forward. Maybe a statement from Topaz saying they are going to concentrate purely on bugs and performance for the next few months would be good?

I’m too old and cynical to be part of a happy-clappy group that will never criticise a product when its blatantly lacking. I have learnt it’s pointless to have devotion for any company. I’ve had my fair share of critiques in my career and I took them on the chin and fixed or improved things. I didn’t just say “Oh you are being a nasty man and hurt my feelings!” which seems to be the go to defence for companies today.

If Topaz steps up then I’ll gladly say thank you and buy another year of updates. That all my part of the bargain entails.

By the way I’ve been using Topaz products almost since they came out but recently it just got so bad I felt I had to speak up. Something has gone badly wrong.


No changes to the Theia model inputs since last version. Refer to my post in 3.0.5 for details.

I hope the dev will not be mad that i quote a comment left on the beta thread but here it is :


Perhaps this link can clarify what development capacity TOPAZLABS have:

Topaz Labs: Contact Details and Business Profile.

1 Like

You can set number of processes to as many as you like, depending on how robust your system is. For example, I’ll use four processes at a time, outputting 3 videos simultaneously while previewing one more, with no trouble. The speed will vary, of course, depending on your system, and it’s dependent more on your GPU than CPU. If you have a relatively weak system (relative to say, rtx4090), you might want to keep number of processes to 2 or 3 at most. Of course, number of processes you set has no effect if you’re only doing one task at a time.

I’m not sure what the reasoning behind disabling autoplay in preview is, but it’s very annoying to have to press the play button every time we’re doing a preview. Can we at least get “autoplay in preview” as an option under preferences?



Marty, I know it’s not the developers, but the people making the wrong decisions need to hear some feedback from users/customers, don’t you think? I want to give you an example. I know a small audio development company that ran into the same kind of trouble, they released software that wasn’t ready. They finally gave the customers that purchased the unfinished software an extra 3 months of updates, but only after they saw a lot of complaints on their user forum. It might be something worthwhile for Topaz to consider.