(hate to bug around but is the VFR bug fully fixed by now?)
Not in my case, still get those weird VFRs, even from the v6.0.4.0b. I gave up and I will just stick with v5.3.6.
5 Likes
a sad turning point indeed 
I knew it! I can’t decide whether this is funny or sad.
No, 29.97 fps → 60 fps resulted in 59,94 fps. 
Hi, i want get access to Starlight
It’s a tragedy, but they have Starlight now and may have other customers then, such as the film industry, and then they won’t need us any more. I hope not and they think of us.
3 Likes
You’re right, we paid for the “evolution”, and now they focus Mickey.
Dream on that last sentence.
2 Likes
I definitely recommend reading the many comments where we’ve confirmed that desktop models are not going away. We are proud to offer a local, offline AI application in a space that is so focused on cloud and enterprise.
They aren’t going away they will simply be smoothly abandoned as mGPU was.
Sorry to say, but many of us don’t believe your word anymore.
To me you simply chose a new business model for the simple reason that TVAI is hacked and massively used as this. We paid a license to enhance “the future”, we have been simply thrown away
4 Likes
Why did you introduce HLG when HDR10 is not even complete, if not to satisfy a future need of streaming platforms like Netflix, Mickey and Amazon??
Who does HLG in your current clients? Nobody …
We have to calculate the MaxFAL and MaxCLL values by hand to get a compliant HDR10 file and not just a “wide gammut” tinkered, while it’s child’s play when encoding since each image is analyzed. It’s really crazy !
You could have even made HDR10+ by introducing dynamic values since the model parses everything anyway, but no … another half-baked thing
5 Likes
This just is not true, HLG is a feature we added at request of users
Ok I may admit that HLG was ask to you because I don’t have access to direct ask and do not browse all topics on the forums, but it’s very surprising that you focused on it before achieving properly HDR.
You know I was to code a tool that make Hyperion output to something completely HDR10 compliant, I stopped after hundreds of hour, disappointed.
I looked towards ffmpeg source code, there are inside is so many strange things you did with.
You removed many things that stabilize the encoding and useful option
As an example, the last I saw if downscale filter. You forced it with Lanczos instead of allowing all the original filters, why ?!? This will some much more added value
There are incredibly inconsistent choices that go beyond the graphical interface that I can’t explain and that would transcend your existing models.
1 Like
Since the video screens are useless when in the ‘Export Que’, the export que should be fully visible when you go to it. I understand you can pull up the Export Que in order to see everything, but it gets old when maneuvering back and forth from different modes. Every time you go back to the Export Que, the default puts it back into a state where you can not see all of your videos unless you ‘pull up’ that screen. Can this be fixed in the future? No biggie, but it would just be more convenient.
Update: It looks like this has been fixed in version 6.1.2 (Mar6,2025). Thanks!!
I DO like the new Live Render. Thanks for upgrading that!
Also, there was a warning when closing an ongoing Export. It is no longer there. If you are in the middle of exporting and you close the video in the Source area, you lose all of your work on that particular video that is being exported.
I have read the comments and I would believe, but what I see happens with TVAI speaks a other language.
Might be a dumb question but does bottlenecking make a huge impact on performance since video ai mostly uses cuda cores and not much of cpu? I have an old cpu and recently got a 4070 Ti 16GB from a 1080 Ti for my second build. There is a 33% bottleneck, would it make much difference since it isn’t gaming usage but encoding video usage.
Not sure if this would help but I had major bottlenecking when I first started using TVAI. My problem was solved by upgrading my RAM, upgrading SSD and installing fans to cool the unit down.
1 Like
Definitely intense on ram. I’ve seen read/write values of up to 60-70GB/s, with average being around 35-40GB/s.
And I believe scaling is done by the CPU, at least judging by the ffmpeg command.
Tensor cores also play a critical part, not only cuda cores.
2 Likes
Thank you, luckily enough I got a lot of RAM and turning on GPU Scheduling allows Topaz to use more of it.
1 Like
