I was wondering if being able to run these 2 features in one pass could be included. Like waifu2x does, it interpolates frames first and then upscales each frame. It would be great if Topaz could include that in VEAI so I don’t have to do 2 separate jobs.
The thing is, it would take just as long since it would still be applying the two models. Each model is good at its own thing as well, and people get better results doing the FPS boosting before, sometimes after, it depends. It’s good to have the choice there. Essentially, I don’t believe there would be any benefit unfortunately. Nice idea though!
IMHO, any serious preprocessing is best done in QTGMC.
1 Like
I disagree, I much prefer Topaz to fix deinterlacing and frame rate
Why, QTGMC is very hard to beat. 
It doubles the framerate (default), and especially shines at its subtle, but very effective, repair options.
Nevertheless, since VEAI does deinterlacing too, it’s not unreasonable to say it behooves them to do it properly while they’re at it.
2 Likes
I found that on version 2.10 at least. Upscaling and using a model takes less time to process 1 frame Than retaining original AR + using a model. 1-3sec/frame vs 8-30sec/frame.
Why?
hi, yes at first i liked dione tv and dione dv because they deinterlace very well, no comb effect, but it accentuates artifacts in places, and reduces them in other places, so since i I discovered qtgmc later, I use this one because there are many more options and the video with qtgmc comes out much more beautiful than with a dione model.
the sole reason for this feature would be to reduce the time and steps needed to do a batch run of videos.
Doing both at once, yes would take 2x as long, but I leave it running overnight, and it would be nice to have it all done on one 8 hour session instead of 2 separate 4 hour sessions. Less output files to keep track of…
I used to use Handbrake to de-interlace using Yadif/BOB but moving to QTGMC (Staxrip) was a major improvement although takes a LOT longer on Placebo mode!
I think this should be a two pass operation, first the upscale and then the interpolation. I think another possibility would be to add an option to run the interpolation model on the CPU and the upscaling model on the GPU and read/write the video as it encodes for maximum efficiency.
