I would say both the nvidia and topaz icons are not quite there yet that is windows 11 ready compared to edge chrome vlc icon
I think the icon change is a bug and will be solved in the next version true the blue lavender one looks the correct icon the green one seems to be delivered with the new menu container from the company that delivers menu packages off the shelf
1 Like
I agree about VHS. I have personal “home movies” from every era: 8 mm film, Super 8, VHS, 8 mm video, Digital8 (480i), DV (480P), HDV (1440 x 1080), and then it is 4K from there.
The 8 mm film scanned at high res, and with Topaz enhancement looks great. But the VHS is very bad, and I don’t really have the workflow figured out yet.
I don’t know how the models are created, but to me, I hope Topaz (or anybody for that matter) will make the effort to train specifically on VHS transfers. Acquire some VHS cameras and tapes, and film a variety of scenes, indoor, outdoor, sunny, cloudy, close up, zoomed out etc. And simultaneously capture identically framed scene on an 8K professional camera. And then train on the data. That would be best, but expensive.
Might also be able to use existing high quality 4 or 8 K video, and then use an optical telecine like process to capture it via VHS era video camera. Should be fairly similar to footage filmed directly in VHS.
And while it is possible that an algorithm could simulate VHS, but don’t think that would work well, as it wouldn’t go through the actual underlying process of light passing through Bayer filter, hitting the CCD, and errors from the writing to tape, and back out. Not my area of expertise, so maybe the underlying analog process can be physically emulted at the lowest level. And you never know, maybe simple simulated footage will have excellent results.
Anyhow, yes there are lots of us with a fair number of VHS tapes, and sadly it is quite bad … before and after Video AI enhancement. The exception being the close up footage you mention.
And my Digital8 footage isn’t any better, maybe worse, but that is another story. I have tried many workflow procedures and settings, but still looking.
2 Likes
The biggest issue I find in converted analog video (not just VHS) is chroma shifting and chroma noise. These are what cause analog videos to have colors that seem to bloom and sometimes look as if they were shot under flourescent lights. There are filters that address them in apps like avisynth and virtual dub, and my guess is that you’ll also find something equivalent in Premiere and Resolve. Once you fix them, you can use most any video editor to correct color balance and saturation issues. None of these things are addressed in TVAI, so you should try to fix them before trying to upsample or enhance.
2 Likes
I don’t get the problem that people seem to have with the greyed out “not processed” view. In most cases, it’s going to be the same view as the original view, and having two original views side by side will just be asking for confusion when someone mistakes one of them for an enhanced preview. “Not processed” makes it unmistakably clear that nothing has been done to that view.
2 Likes
windows 2000, worst edition! lol
1 Like
You know that you can make either side be original or preview/not processed, right?
1 Like
If you use stabilization there´s no pause function.
In this case you have to only stabalize the video first and load the exported video and use the preset again this time disabling stabilization…and now it works.
I stabalize all my videos now first before doing anything else.
I’ve had mini VHS tapes upscaled.
It depends on how you transfer them as this can cause a loss of quality too, are you transferring them from Composite or S-Video input? Have you thought about investing in a RetroTink 4K upscaler? I currently own one, they are expensive at $750 but worth it. What I’ve done is transfer it using the best input available which for VHS is S-Video or HDMI if your device has it but doubtful due it’s age.
Then send the signal from the RetroTink to your computer’s capture device, in my case I have a Elgato 4K60 Pro, then capture it in 4K then downsample it to 1080p using Handbrake in lossless settings then feed it through Topaz at least 3-4 times, best results by upscaling to 4K then downsampling it back to 1080p each time you upscale it.
My videos from my grandparents old vhs camcorder turned out pretty good, not close to native 4k but does a good job with minimal distortion. Sadly videos below 720p with low quality take a lot of time to get any decent results, it’s a long process and sometimes requires additional equipment in preserving the image quality when transferring them to digital format.
i dont have windows 11 so im good
The newest of my workstations runs W11. I don’t notice any difference in performance compared to W10. It’s the Pro for Workstation version, so that may make a difference.
no? rgb and yuv444 allows odd resolutions i litteraly said it right there tf
When saving projects to a custom folder manually (I have the default path set but wish to save a specific project somewhere else), the previews are always kept in the default projects folder path.
Related but has been a thing for a while, the program doesn’t ‘‘let go’’ of preview files and they cant be altered as they’re in use according to Windows. TVAI needs to be closed entirely for files to be moved around, previews and source.
I prefer to stay on 4.1 atm
- Performance is same (4090, 7950, Win10)
- I prefer temp output file in source dir
- New interface is ugly in parts (esp colors)
- No new interesting features for me personally (pause- resume i do manually)
1 Like
I feel the same way (both for Video AI and Photo AI). It amazes me that the UI changes are (probably not only for me) so relatively large. I find the direction of Topaz software development somewhat oscillating and although I paid (perhaps unnecessarily, I’m afraid) for another year of updates earlier this year, I don’t dare install the constant updates just to reinstall the older ones again.
The focus of Topaz tools seemed attractive to me about two or three years ago, but I have stopped understanding it now. Based on the progress so far, I expect it to continue like this, so I’ll have to find something else. It is too expensive for the offered experimentation and dealing with changes of the UI type. The use of AI for video and photography is booming at the moment, so there will certainly be alternative paths and Topaz will succeed with its approach easily without me. Nothing special, sometimes it’s like that with some sw tools – it’s just that I don’t want to risk money for something I won’t use. For photo and video professionals it’s certainly much more fundamental; for me it’s just fun, but I’d still like to use sw tools with a more stable development path.
6 Likes
aion still fails, any news on a potential fix? limiting vram doesn’t fix it at all
2024-04-14-16-29-10-Main.tzlog (314.1 KB)
I did notice “penCV(4.5.6) Error: Assertion failed (u->refcount == 0) in cv::StdMatAllocator::deallocate, file C:\Users\Gregory.conan2\p\b\opencb7b82bbe96ee7\b\src\modules\core\src\matrix.cpp, line 169” there is no user gregory on my pc, I’m probably just misunderstanding what this refers to, though.
I just upgraded to 5.0.1
It had been awhile since I used Video AI.
The interface seems different and I don’t know
what I need to do differently.
Is previewing what I let run until exporting?
Current 1:39:00 video is taking 22+ hours on my 32GB M1 MacBook Pro
It would be nice if you could post a tutorial video on your youtube channel.
Thanks
1 Like
3 Likes
Does the ofx version work in Fusion Studio? Tried 4.1 and 5.0.1 and they both throw up this error.
Error: com.TopazLabs.VideoEnhance returned kOfxStatErrMissingHostFeature from OfxImageEffectActionRender
EnhanceAlpha1 failed at time 0
