Topaz Video AI v4.0.6

I have both versions, running quick tests on 4 betas but when I need to do a real job, I do it on v3. I still love the product, especially the most recent Iris model, but v4 is not ready for prime time for me.

2 things I am eagerly waiting for: live previews on v4 and (much) faster Iris on Apple Silicon.

Thank you for the hard work!

1 Like

Can anyone tell me exactly what ‘old’ means in the context of the message “Remove old model files?” during an install/upgrade to TVAI? If Topaz have provided details of this previously, a link to it would be very helpful. I always select “no” simply because I don’t know what ‘old’ means in this context.

I downloaded the new updated upscale and now it’s trying to charge me. I have had this app since April 2023 and just finally been able to use it the past week and it keeps failing. Can I get some type of refund and a restart?

As they make little or big changes to the AI models, they put them on their servers for download, and the program stops using the old versions on your storage drive. They are just taking up space at that point.

1 Like

And sometimes it’s good to still have the old models around if they did something like with Iris on the Mac, extremely hampering performance in the newer ones…

1 Like

I really wonder what they’re working on that’s distracting them from reverting that. I’ll bet it only takes ten minutes to make the fix to use the old model again.

Will the upcoming Resolve OFX plugin support Linux? (PRETTY PRETTY PLEASE!)

where do i download the linux beta?

I think you have to be a beta tester. And now I realizez since I haven’t posted for a long time in the beta forum it seems I have been excluded from the Video AI beta group :grimacing:
But still doesn’t make sense since I still see users that haven’t been seen for more than 2 years ??
My last post was maybe before this summer.

1 Like

Just updated from Major version 3, I’m used to looping a 2 second preview. What’s the best practice for me to train my eyes on the same processed footage?

Most info space is given to the Previews by default where it is not needed while the Exports infos are cut off at the end.

well went back to 354 also from reading on here and have to say it works well for deinterlace and vhs. they should combine 3 and 4, the preview for sure is better easier to see which models are effective for you,my 2 cents hope they get this right soon

1 Like

Well, I’ve decided, folks. From reading from here and from there, it’s clear to me now that TVAI is nowhere near ready for anything serious, reconstruction wise, of any videos - personal, enthusiast and definitely not anything commercial.

I mean… from the oodles of complaints about monster’s faces (it’s even generated on non-human objects like vases, ect.) STILL not corrected for almost a year, no bt.2020 colour grading, no colour correction, the preview player is possessed for all versions, simple FULL metadata statics are MIA, roadmaps never gets TOTALLY fulfilled, clearly hardware acceleration is extremely limited and the claimed of the use of TensorRT is mostly only marketing in my opinion, by just going off of many metrics - on & on.


Stop spamming the forums with this. Any future posts of it will be flagged and removed.



Well, but since this is an ongoing issue I believe that one post about this for every release should definitely be OK!

First, you can really argue if you absolutely need AVX2 CPU support for an app that should rely mostly on GPU.
Secondly, and even more important those that managed to do a manual install of TVAI on non AVX2 CPUs report that it does in fact fully work and not even slower - so this limitation seems very arbitrary at least for the moment.


The use of TensorRT increase the speed at around 50% on Nvidia GPU (RTX 40 series)
(Assume your system is not bottlenecked by CPU).
You can compare the speed of Iris-v1 (LQ) without TensorRT vs Iris-V2 (MQ) with TensorRT.


You can buy new PC every month - who cares and who did ask you about this ?


So as a “Leader” here you call your customers kind requests “SPAM” instead to do everything possible to help them ?!? I just ask for help and ask you to bring back something that already been worked fine !Nothing more ! I say PLEASE !
Honestly I believe with such “Leaders” - this great project is doomed and this is very unfortunate …

1 Like

I would do but not every month better hardware is released but probably i have some second hand hardware that you can purchase which better supports your Topaz software. I really like you lemans is not formula 1 the latter is really faster


Hi guys! Is the ‘wiew certain part’ on previw loop fixed?