After a remark of @wendell in the Level 1 News about an up scaled version of Deep Space Nine to 1080p I have looked into the subject a little.
I found a tutorial using Gigapixel AI. Essentially you extract every frame of the video to a single image. Each image is then run through Gigapixel AI and merged back together after. That is quite some work, especially since all the individual frames have to be added to the program manually (and you can’t do all at once unless you have like 128GB RAM or something) and no command line interface.
The same company also offer Video Enhance AI. I have to assume it does something like that without any additional work from you and you can process multiple videos at once. The major downside of that approach is the price of the software. They are asking from 300$ for it.
They do offer a trial and I have played with it. At least to my eyes the result looks good. I have up scaled the first 3 episode of Star Trek Voyager. I do own the DVD and have up scaled to 1080p. They do look okay, of course not like native 1080p.
Has anyone else tried anything like that? Any alternatives to this software?
@wendell as a major Star Trek fan, if you are interested I can probably get you a few samples of the up scaled results for your judgment.
Most of the AI upscalers for video will process video so you do not have to extract frame by frame, but there is also an option to output in image sets to have higher quality output - 8 bit png or 16 bit tiff, if you are a quality snob and have the space, time, know how, to deal with that sort of thing.
Video Enhance AI can be CLI’d to pass through AVI scripting to make it semi automated, I cannot speak to the other programs that do this sort of work
I know “there are no stupid questions”, but this still feels awfully stupid to ask: How? If it is on the forum I don’t seem to find it. If you have another preference, please let me know.
I just picked up a He-Man and the Masters of the Universe on DVD and while the original video quality isn’t totally garbage (480p IIRC), I’d be interested to see if AI could upscale it and do a decent job.
Right now the results are promising. I do it on the Voyager series. And you can get a 30 day trial for the software.
My workflow right now looks as follows:
Pull the DVD files with MakeMKV into an MKV file.
Process the MKV
Video Quality Low
Video Type Interlaced
Video Artifact Type None
Sizing Custom Setting, Scale 187,5 with Lock Output Size
Compression Facotr is 0 (Waring really big file size)
And keep Audio to no
The process takes about 2,5h for 45min of a film. I run it with a 2080 Super (and yes, the graphics card is required otherwise it will take 24h and more)
Remux with MKVtoolnix the original audio and upscaled video stream into a new file
Process with Handbrake and the Apple 4k preset (which mean x265 encoding)
You can probably set keep audio to yes and use a higher compression ratio if you are okay with the H.264 encoding, but for me I process it afterwards anyway so I want as little compression as possible.
Also the on these AI settings, the intelaced video is deinterlaced and actually doubles in frame rate. I do not see any major drawback in that, since my output size is okay for 1080p on H.265 but some may not like the taste of it and I do not know what the audio does if it is processed by Enhance AI directly.
I do not have any issues with it. But also a remux the files after the fact.