In VEAI 3.0,0,6.b it is possible to deinterlace without enhancement. That should be pass one.
Pass two should be to resize and enhance.
However, VEAI deinterlace is not yet up to snuff, especially when it comes to highly compressed input.
If you can do the Deinterlace in a solid 3rd-party application, If you need one that does high standards work, I suggest you look into a program like Acrovid FootageStudio or one of the opensource applications,
The reason I emphasize high standards is to avoid GIGO. (Garbage in, garbage out.)
Testers of VEAI know the Model performance very well.
I would like to see lists of the models/json version used by VAI 3, going forward.
Company data around GPU/CPU variations, and explanation of the different sizes used?
Although I understand the logic to sell all features, I would like to control exactly what models are downloaded. To this end I need a transparent explanation of what models are needed for me.
Regarding ghq and proteus:
If all usable models are installed, the app should not be attempting to down load in case something is updated. What I am purposing is the app is controlled by me, not the other way around.
At this stage I am going to keep an eye on others feedback for performance information.
Not keen on installing a program that can only be used when dongle is in internet range.
Like to see 32bit Lagarith output option on UI.
I just wanted to add my opinion on things.
Cheers
In version 2.6, I used Proteus for many PAL DVDs without first eliminating interlacing. And it did a great job.
Is Proteus as good in version 3? Only those who have done it themselves (PAL DVD - Proteus) or have accurate knowledge are welcome to answer.
@ida.topazlabs@suraj One thing which would speed up things a lot for my usecase, would be the possibility of scaling straight to 1.5x with the AI filter itself.
I have a folder with about a hundred 4096x4096/3840x3840/3840x2560px stereoscopic 360 degree-videos, and upscale them to double the size, but unfortunately after that I have to shrink them under the max of 35 million pixels (with the scale_qsv filter) which is the maximum the HW H265 decoder of the Meta Quest 2 can handle. It would speed things up a lot to just upscale 1.5x. Would that be possible?
Exporting to h264 would be possible but would either be a bit blurry or stutter because of the high bitrate demandâŠ
Last night, I imported a Quicktime (.mov) source file created using Avid DNxHR HQX.
I tried this because I was informed that it was a nearly lossless format and could go through multiple exiting/filtering processing passes without suffering any significant image degradation.
I played the original file in VEAI. Other than being in desperate need of resizing and enhancement, it played but I could not generate a preview,It failed.
FYI: I have selected .mp4 as my default output format.I changed the output to .mov and then preview worked correctly
Since I rarely work with Quicktime, I donât know whether this behavior is something I should expect or is something that needs to be fixedâŠ
@ida.topazlabs I still find that my XAVC-L based MXF and MOV files play back at about 2-3fps on my M1 MAX 32GB. Also, previews play back at about that speed too. This makes judging things like stabilization very difficult.
@ida.topazlabs Pressing the âFitâ button on the zoom control causes the video to get considerably smaller than the available space in the window. For exampleâŠI press it and scales to 24%âŠthen I manually scale it up to 28% to get it to fill the window. This is a 4k video, so 24 to 28 is a big jump in size.
Attaching an example of the âfitâ issue. Behavior in most apps is for the âfitâ button to maximize the video size to the available window real estate. Always want to see as much detail as possible.
@ida.topazlabs One more thing. The messages given when a preview is rendering are not very descriptive and quite confusing. Iâve attached a video (slowed down to make it easier to read)âŠwith some comments showing what goes through my head as I watch the status area/progress bar.
I just quit and ran it again. This time I did not get an error, but the app just hung on âRunning Model,â and ffmpeg is not running in the background. Attaching logs. 2022-09-16-11-32-45-Main.tzlog (945.0 KB)
@ida.topazlabs I quit the app and ran it again. Same settings Stabilization+ProteusâŠThis time it ran, but the results were crazy! Attaching a video. Thank you for all of your hard work!
@ida.topazlabs Donât want to overload youâŠbut this was a good day for me to do some QA
When I drag-and-drop a file into the app, the behavior is peculiar. First, my computer âthrows it backâ indicating it was not accepted. ThenâŠI waitâŠ1âŠ2âŠ3âŠNearly 3 seconds later the video does actually load. The first few times this happened, I thought it wasnât working, so I re-loaded the video. The initial behavior is oddâŠbut the time required to acknowledge loading in the UI is the real issue. Attaching a video.
