After using numerous video encoders over the last 25 years, in what has seemed like an endless pursuit to improve the quality of my beloved 80’s music videos, I decided to go with Topaz Video AI with the hope this might be the last time I have to start from square one. I was already a few hundred videos into the process using Vidmore when I heard about Topaz Video AI so I stopped where I was. I have all my DVDs ripped into individual vob files and have well over 1,600. I have quite a few more I couldn’t find on DVD, from YouTube and capturing from VH1.
I searched the threads here, but didn’t find any pertaining just to music videos. I’m curious if anyone here uses Video AI to enhance/restore music videos. I do have a few questions about this software.
Due to the fact there are no sound/volume controls in Video AI, does anyone know of an app that can “normalize” the sound level of a vob file without having to re-author the video? When it comes to compilation DVDs, the volume levels vary from video to video.
Can someone suggest which settings would be best for music videos? It seems like every YouTube video I watched suggested a different model.
Is increasing the fps to 60 recommended or not?
Given that most of the videos are 480p; should I stick with upscaling to 1080p, or can Video AI do a good job upscaling all the way to 4K?
Is there a way of estimating a processed video’s size, given which settings are used? A user created calculator?
I appreciate any/all advice and comments. If anyone is interested in trading videos, let me know. I’m always interested in increasing my collection.
Seems like a good community here from reading posts. Looking forward to learning more about Video AI.
VOB files makes me think the encryption may not have been removed from several of your videos. TVAI won’t be able to read those.
I thought that the mkv container had some sort of replay gain type feature, but it does not. You’ll need to re-encode any audio that’s not the same standard level. I use GoldWave to set the volume to LUFS -23dB. Since you mention Vidmore, with that, you should be able to convert back to AC-3, or DTS if you want more bit rate. (If your setup can do flac, maybe just stick with that. It’s smaller than DTS master audio and lossless.) Either way, the mkv container makes it ultra easy to join that modified audio back into the final video without doing any conversions.
Don’t bother with 60 FPS at this time. It adds a layer of complexity that is easy to get wrong. You can always do it later once you’ve got everything upscaled. I recommend upscaling before interpolating anyway.
I’d stick with 1080p. You are less likely to get good results going all the way to 4K.
For models: I avoid Iris and Theia. My go-to is Proteus. If the movie is super clean, like low grain, low DVD compression, Gaia is great for that. That’s usually only CG content. I’m still learning the tricks of Proteus 4, otherwise I would have better recommendations for it. For the last full TV series I worked on, I did Proteus Manual:
compression=0.4 (That's 40 in the GUI)
details=0.12
blur=0
noise=0
halo=0
preblur=0.2
blend=0.32
It was a show with heavy grain. Maybe these setting will be a good starting point for some of your videos.
As to what the final size is going to be? I output to tiff images and use another version of ffmpeg to use the software encoder to convert to H.265. At CRF 20, the 1080p result is usually smaller than the DVD rip.
Thanks for your reply and suggestion! I tweaked the settings you shared and the results were great. Upscaled to 1080p looks better than the original by far. I then upscaled the 1080p to 4K. The only issue I have with the 4K version is that the blacks/dark colors are kind of pixelated. Is there something in the settings to smooth out blacks/dark colors?
