I have been trialing VEAI and it’s producing some great looking results, however, I’m unsure if it’s the right tool for my particular application. I’m unable to try the 3.x alpha builds without a licence, so the present lack of command line support has made me look elsewhere. I have several thousand short videos, so batch processing is essential.
I have been trying to improve the quality of some VHS captures. These were captured using Blackmagic Design cards and lossless compression from first generation recordings made on professional gear, so they’re probably about as good as they are going to be coming off 30+ year old VHS tapes.
A few years ago I created h.264 versions of the collection using ffmpeg. There remained significant room for improvement and initial experiments with VEAI produced some nice results. This was again improved by using QTGMC before VEAI.
I note, however, that when scaling to 100%, VEAI indicates “Denoise / Deblock”. Does this mean the AI features are disabled and it’s just doing the denoise and deblock? VEAI isn’t messing up fades like Real-ESRGAN is, which suggests it’s either a lot better, or not applying AI.
I would say using QTGMC and enabling denoise/deblock, I’m 90% of where I would like to be compared to VEAI. Should I continue tweaking in Hybrid, or is VEAI adding something extra even at 100% scaling?
