I recently purchased TopazLabs to Upscale and Enhance a feature film project that was shot back in the early ’00’s. After watching the informative video tutorials for Video AI, I was hoping to get a bit more advice.
My film was shot on a Panasonic DVX100a at 720x480 with anamorphic squeeze (it’s essentially 480p non-interlaced video footage). “De-squeezed” results in a resolution of 854x480p.
There might be potential interest from a streamer to license the film, but they would like the resolution increased to a minimum of 1080p FHD and hopefully 4K. I’ve run some preliminary test with the Rhea preset as it is suggested for upscaling footage. The results have been impressive. Given that I also would like to enhance various quality aspects the video as well, but I’m still a little unclear about a good workflow after watching Video AI tutorials.
Some questions I have are should I be using the Rhea preset for the upscaling? And should I do an upscale pass 1st and then do a Second Enhancement pass for various other improvements or vice versa and target quality enhancements 1st before upscaling. Or, do both at the same time?
The tutorials also mentioned tackling standard def VHS and VHSC video footage with Nyx, Iris and Proteus. Just hoping to get some insight on what presets one might gravitate towards given my footage parameters?
I know it can be very subjective and every film and shots are different. I’m just hoping for a little more advise or suggestions on a particularly workflow given my footage circumstances.
Hi there, every has his own workflow and hundert meanings. What works good for me is
deinterlace first when its interlaced, I do it into Hybrid with QTGMC, settings pending on noise that should be removed. Keeping max details is choose preset “very/super fast” or “draft”, as slower you do more noise gets removed (but also some loss of details), export it when you have enough space with lossless codec FFV1
When you have 16:9 DVD PAL content, De-squeeze to 864x486 instead 854x480. 864x486 is exactly 16:9 and codec friendly, produces no floating points. I do this steps in Virtualdub2 because of visual check when also crop is needed, but you can also do everything into Hybrid
Simple Upscale
Chosse model of your choice, try out different models and settings. I recommend for most models try first just Deblur, ok sharpen and Rec Orig Details are also allowed, but take care, often users does to much and some functions are very invasive
2-Layer method
Made two upscales, one with Rhea or Proteus as base layer and second one with Iris MQ or LQ, then overlay booth frame exactly in your video editing software and set opacity for example to 50%, so both vids melts into one. Optional: import the merged result into TVAI and do some finishing steps, I often use Artemis “Medium Halo” + “HQ12” or just Artemis HQ12
Rhea Base Layer method
I recommend try this out, it’s easy and can give good results. Use Rhea auto mode without user settings 1x (no upscale). After that import the result into TVAI, set desired upscale resolution and model “Proteus” in auto mode often without user settings (you can also turn Rec Orig Details to zero). Proteus V3 is for keeping max details or choose Proteus V4 when video is still too noisy. Thats it.
