Absolutely! We have samples going out across our social channels over the next few weeks.
@ForSerious MPEG-2 should be fine. Could you click the red error and paste the details?
If your red error text doesn’t show a pop-up when clicking, send me an app log or cloud request ID please!
Hi Axel, this is a good note for us since we actually are trying to have the plugin not load in Fusion to avoid crashes.
Our engineering team will look into this.
This specific 6.1 build is 6.0.4 + Starlight, so yes. But in general, we fully expect to keep updating Video AI in major ways for desktop users.
1 Like
Hope you had a chance to try it out with Starlight! From the screenshot, it looks promising. Similar quality to some of our validation videos.
Invalid input
That’s all there is. I can send the logs if you put up your drop-box link.
I get the error “Couldn’t create request - Diffusion models currently support videos between 5 and 300 frames.”
Original is 24fps, which at 10s is 240 frames. The original is 14s long, which puts it just above 300fps, but that shouldn’t matter since it automatically cuts it down to 10s?
2 Likes
This is definitely very good news to have as early as today: let’s say that since the FOCUS of the program is (or should be, or at any rate has always been until now) LOCAL rendering (cloud rendering MUST remain one more option)… by releasing this new model ONLY with CLOUD rendering, let’s say that you have broken a tradition; this is what created for me a bit more concern… and I believe that this new model should have come FIRST as a local rendering version and only LATER as a CLOUD version, or at least make it available right away in both ways!
The concern remains that: YOU SHOULD NOT shift the focus to CLOUD rendering, because at that point there are many basic FREE online tools where you ONLY pay task credits… in this case we are talking about a PC program that works mostly locally and costs $300 + $150 on average each year… I would say there is and should be no comparison!
14 Likes
The count is based on frames so as long as you’ve selected an area under 300f, the process should begin.
Could you try a range just under 10s to see if we have an issue with the counter?
I think this is a viewpoint that can be related to most technology. For personal computers to advance into sleek, energy-efficient devices, quite a few large power-hungry computers had to be developed and iterated on.
If we only ever worked on models or ideas for improving quality that could immediately run on consumer hardware, we’d make progress much more slowly.
Require internet connection to use local software. Software releases start to fail to remember auth status.
Feature-gating with TVAI Studio, removing functionality from the prosumer proposition.
Annual software subscription operating like a monthly-fee in all but name.
Cloud pay-to-play option.
Cloud-only models.
TVAI’s USP was “local processing with a convenient FFmpeg-friendly, scriptable command line”.
If cloud is the future of TVAI, then the future looks bleak for existing users.
12 Likes
Does it not accept 59.54fps (What VHS correctly deinterlaces to)?
I deinteralced a clip and saved it in H265 and tried that, but get the same invalid input error.
I even shortened the clip to like 2 seconds. Same error.
Welp, I give up. I’ve tried five very different sources and they all end with the same invalid input error.
1 Like
I do not question technological progress: it is obvious that with CLOUD rendering performed by a significantly more powerful PC, better results can be achieved…
…I do, however, question the “business model” of the program: as already pointed out, those who paid so many dollars for the program did so in order to have RENDERING IN LOCAL with their PCs, without the need for an internet connection and ESPECIALLY in order not to pay the “tax” of cloud rendering for EVERY freaking video: the fact that there is, in such a program, may be something “more”, but it has to be just that…
if from one moment to another, an upscaling model ONLY with cloud rendering comes out, then as users we are allowed to “scare” ourselves, because the direction taken seems to be another than the initial focus!
12 Likes
So what happens when one of us in a state with looser restrictions cloud processes something that results in a shareable link on your servers that could get you folks in TX arrested on obscenity charges…?
2 Likes
Agreed. I am extremely disappointed by the decision to make this cloud based 
6 Likes
Second this.
But, unfortunately, the shift to the cloud is already in full progress: look at Gigapixel and PhotoAI or the Gigapixel iOS app - the “force” to use the cloud is drastically increasing.
What’s even worse is that Topaz seemingly are (are least currently) not ready on their server side to withstand the load they cause with that approach - looking at sloooooow cloud renders already in Gigapixel and Gigapixel iOS with partly fully hanging queues and encodes that never get finished (but for which you get credits withdrawn) I don’t even want to imagine what’ll happen with generative AI video processing in the cloud…
6 Likes
Cloud only? Starlight does not approve.
9 Likes
So the statement is that the customers’s hardware is to slow for the fast progress that Topaz Labs wants to make and the cloud is the game changer and makes the progress possible.
ok but then I ask myself why a 10s test-video in the cloud takes up to 20 minutes? Then I hear statements the cloud generally was slow. I have nothing against gpu hungry Starlight model renders locally on my computer that is as slow as your cloud
latest technology is in my computer, called NVidia and AMD and my computer can run latest program code!
Maybe I’m just too critical and everything will be great, I am silent now and observe.
2 Likes
