I go through three steps to upscale to HD from DVD. I’d like an option to do three steps as i find I get the best quality this way. I normally use the Artemis Med model for DVDs.
For a 4:3 movie, these are the steps I go through:
Upscale to square pixels: 720x540
Run it through again and upscale 1.5x to 1080x810
Then run it through a final time and upscale 1440x1080.
16:9 DVDs go through 854x480, 1280x720, then to 1920x1080.
It would nice to be able to program in all three steps from the start instead of running the same video through TVAI three times.
Another feature I’d like to see is the ability to stop the queue processing like Handbrake does. It asks you if you want to complete the current job then stop or just stop now and safe the job’s progress (if possible), or abort job if it’s not possible.
Also when starting TVAI it would be nice to tell TVAI to start processing the queue instead of just automatically starting the queue on startup.
I never tryed out three upscale steps, but it’s interesting. I think when it’s usefull, then when you go higher than 1080p. DVD resolution is 720x576 and when you do 720x540 you lowering quality, scale to 768x576 is better.
For 16:9 854x480 is bad, because it’s not codec friendly, gives floating points and it’s not exactly 16:9. You can scale 720x576 DVD content that is made for 16:9 play to 864 x 486 then import this into TVAI. Or when you have already 854x480 source, you can leave it and put letterbox around it fits 864x486, then crop the little letterbox away on upscaled result.
Two times Upscales I did already. This can make sense when footage is hard to bring it sharp at a wanted high resolution, when artefacts happens. Then you can do for example with Artemis, Proteus or Iris from SD to 720p or 1080p and from there to a higher resolution. This could allow going from SD to 4k if the source normally doesn’t allow without artifacts. The disadvantage is perhaps the naturalness can suffer when you use two or more times reconstruction models for upscaling process. Textures can be smoothed too much, of course it depends on choosen models and settings.
I supposed I should have mentioned I’m dealing with NTSC DVDs and not PAL. So those resolution steps are for NTSC DVDs. For PAL the first step would be 768x576.
I notice a nice uptick in quality with with everything. But it is especially noticeable with hair and clothing textures.
I do preprocess the video through Ripbot264 to deinterlace/de-telecine to 23.976 fps progressive (if that is the original source), degrain using SMDegrain, but I keep output to still to anamorphic as the first step in TVAI is to convert from rectangular to square pixels and I find this step to be critical to to getting the best quality.
the more you’re trying to achieve, the more complicated your working queue will become.
and at some point it doesn’t make sense to limit yourself on doing all of those steps in a single program with a very limited scope of tools.
topaz vai will never be the best at deinterlacing, fixing colors, fixing borders, getting rid of temporal noise, getting rid of banding, it cannot splice the best parts of different model encodes into each others etc ppp… this is not its core competency.
you’ll waste lots of time and opportunity to force yourself and the program to do something which can be done better elsewhere.
I’m using TVAI for the upscaling using the Artemis model. I’m not asking it to do the preprocessing that I’m doing before adding it to TVAI. I was just mentioning that I took care of all of that before adding it to TVAI.
I pretty much just would like to see a feature that allows multi-step upscaling as that will speed up my workflow without having to run the video through TVAI three times.
