I bought a licence over a year ago. It’s “upgrade” term is over and I didn’t bought “upgrade” yet. Because it’s mostly not upgrade but “update”, which includes many fixes. An upgrade must looks like a new model for example “starlight project”. But it comes with server thing, which requires additional payment. That means we have to pay additional of additional money for an “upgrade”.
My answer is NO. I will look for another software.
by the way I have licence for sale (without “upgrade” term)
Uhm… c’mon, Eric!
NVIDIA’s says: “Say hello to my little friend” - RTX Neural Texture Compression!
~1/8 of the memory with Nvidia RTX Neural Texture Compression e.g. H100 80VRAM - but with Nvidia RTX Neural Texture Compression that footprint drops to ~8.5-10VRAM that in theory, a 4060/5060’s could potential run these “server-grade ONLY” models. 
The huge size of model? I have terrabytes of ssd’s, waiting the local release :c
And also some of us have 5090
This model does not run on the 5090 in its current form
Will me having upgraded the license today even have any significance to this? Or do we need to pay per video/minute here in a way that’s disconnected from the desktop app?
Maybe you can post the Demos on your Youtube channel as well.
lright, let’s dive in. My name’s Roberts, and I’ve been here since day one—back when Topaz Video AI first launched. I’ve watched, downloaded, and tested every model across every version since the beginning. Over time, Topaz and I have become close companions, and I’ve invested heavily in upgrading my PC (started with Intel HD Graphics, now rocking an RTX 4090) just to optimize performance.
While I truly appreciate the team’s effort to innovate with new AI models, I’m struggling to understand the push toward cloud-based processing. Why not refine this model locally first, then release it once it’s fully optimized for offline use? Privacy has always been a cornerstone of the Topaz community, and the cloud feels like a step in the wrong direction. Let’s keep processing local—where it belongs.
For clarification: Many concerns stem from the lack of transparency around key details. Why not share specifics like model size or the technical challenges of running it locally? Even basic information—such as hardware requirements or computational limitations—could address doubts and foster community acceptance.
We understand that the current iteration might not run even on a “RTX 5090,” and we acknowledge that server-scale resources may be necessary. However, keeping these numbers overly secretive feels counterproductive. Sharing at least some metrics—even if they’re daunting—would demonstrate honesty and help users contextualize the limitations. Transparency builds trust. By openly addressing these questions, the team could turn skepticism into collaboration, aligning the community’s expectations with the project’s ambitions.
I think somewhere Topaz said it currently uses an NVIDA H100s to process the videos.
Topaz Labs thinks people will be excited about Project Starlight. In reality, it’s a wake-up call that Video AI is never going to improve past this point, and all users should start investigating alternatives.
Starlight needs to run on 4090/5090 locally. If it’s slow, then simply provide quality grades between existing models and Starlight. The fact that local models have not increased dramatically in quality since Topaz Video AI’s initial release, combined with this cloud strategy, says to me it will never get better.
After the Adobe debacle, everyone in the creative community is concerned about uploading their footage to the cloud, if they aren’t already contractually prohibited from doing so.
Topaz, give your users choice, or the moment the DeepSeek equivalent of video upscaling hits, we will be gone so fast your heads will spin.
Hard pass until this can run locally.
I pay for, and renew Topaz by evaluating the value and the cost. I do not value cloud models in Topaz, so when renewal time comes around, we’d best see local improvements!
Starlight is an interesting tech demo. It will matter to me as a Topaz value proposition when it can run on even fancy local machines. I imagine folks will tolerate a slow model that does a great job. Hopefully you find a way to fit it in a 24gb card.
So the fact is Starlight is by far the best AI Video Enhancement out of all the current models in Topaz Video AI, I am very impressed with that it can do with VHS footage. That said, I have ZERO interest in any kind of Cloud based service to get these results, as it would cost a small fortune just to render out a 30 minute or 60 minute project…NO THANK YOU.
I know you guys like this idea of doing things in the cloud so you can charge us in an ongoing basis but as you can see by ALL the replies here, no one is going to bite on that. And the fact of the matter is you guys are not the only circus in town on this. There are others out there coming up with the same and similar products for AI video upscaling and in time, they will offer something as good as what you guys have with Starlight for desktop computers on a local level I am sure, it is really only a matter of time.
If you guys are hell bent on keeping this a cloud based service only, good luck to you in that because I will not be on board with that at all, nor is about 90% of the replies I have read on this thread.
I can only hope that Topaz doesn’t become the “Swiss Watch Company” of the modern day era, where once upon a time the Swiss owned the watch making market, but because they failed to embrace change and address what the people wanted, they lost the market to Timex and Casio. You don’t want to become The Swiss Watch Company.
Where is the problem? Many of us have access to H100s and better GPUs. They’re quite cheap to rent nowadays.
That’s what we’re offering with the cloud early access service. You could rent an H100 by the hour, but you’d need to spin up an operating system, drivers, configuration, and initialization each time it is used.
Yes, it would be less convenient and more expensive, but your cloud service would not be logging my video in that case. That is of critical importance to me and I suspect many of your other customers too.
Once upon a time, Topaz you told me to upgrade my video card. So I did. After several versions and asking for money to renew, I believed in the future. I believed in new models and I trusted that the buggy software would be fixed. Then came new GUIs, old bugs, new bugs and dated models. Topaz said I should upgrade my video card…that things would improve, and asked again for renewal money. So I guess by now I’m pretty stupid because Topaz wants me to use their cloud processing. Heck, why not. It seems like sweet deal for a dummy like me.
100 percent agree with @gumba1963
I have paid much for my fancy desktop and will keep it updated so I want it could tacle with hard workloads like upslacing. And if I would pay for “upgrade” which is actually “update” I don’t have pay for new models. In addition even if you give up upgrade payments I will never pay for the cloud thing.
I advise you to think again.
So my specific problem, is that for almost the entire version of 5 of Topaz, the software has been essentially unusable for me despite having paid the renewal to use. I get that you can use any version of the software indefinitely - but the free updates to current version whilst software renewal is live are kind of pointless when it doesn’t work anyway.
As an FYI, I reported the issues multiple times in multiple ways with 0 feedback or assistance.
Below are my posts reporting issues from Version 4, that then persisted into Version 5 and became worse.
Below - after having no luck in forum posts was my message, I tried to message Tony directly - who I see is in this thread but never responded to any of these - and didn’t respond to my direct message either.
So I gave up trying to test the new versions and periodically just checked in to see if the program was usable yet.
Late in version 5 - around 5.4 I believe - you changed the preview render engine and then I was able to load sequences again. Which was great, I had only waited months at that point to even use the application, only to still have multiple issues with the image sequences all of which existed prior, but now with new ones, like cutting off the sequence with no explanation and random chance as to whether it will process the entire video.
I saw when V6 came out to try that one and the same issues persist, so i stopped trying again.
I am not sure exactly how I am to feel about being told that you value the testing and input - for now a cloud based rendering that essentially will need to pay extra for - when I was left hanging with a paid for unworking application for MONTHS without anyone at Topaz having the courtesy to even try and reply, assist or fix the problem.
So apologies for vent, but it would just be nice to at least start with a program that works at all before being asked to test for a higher paid service.
