ireactions wrote:Something I noticed: the Lanczos versions of my upscales are very hard to shrink down from their 4 - 6GB file sizes. Compression reduces the video noise that's alleviating the visual issues.
I was also running my Topaz upscales of THE DEAD ZONE through Handbrake to reduce the file sizes a bit, taking them from 3 - 4 GB a file to about 1 - 2 GB a file. The video quality went from being a high quality film scan to a medium quality 1080p Netflix videostream and some of the film grain texture left intact by Topaz was screened out by Handbrake.
Other than storage capacity savings, what is the primary reason to reduce sizes? My understanding is a blu-ray movie is pretty larger (and that's about 2x length of a sliders episode) so moving to 1 to 2 gigs would seem to be messing with the quality, no?
And you are seeing some of that I guess...
ireactions wrote:Out of curiosity, I ran the very grainy blu-ray Turbine version of "As Time Goes By" through Handbrake at the "Super HQ" setting -- and Handbrake blurred out almost half the film grain and there was a slight loss of video quality. It's interesting: Universal's DVD compression in the early 2000s seemed to add compression noise; Mill Creek compression in 2016 added noise.
However, modern video compression codecs seem to screen out video grain and noise in the process of crunching a video file into a smaller container. I ran "As Time Goes By" but with the file size set to "Production Standard" and this time, all the graininess (and sharpness) remained.
I have to wonder if my technique for processing the DVD files has been a problem with SLIDERS, episodes 102 - 109. While I copy the file off the discs, I have to run them through Handbrake to deinterlace them or the final file has horizontal lines flashing across the screen.
I ran Seasons 2 - 5 through this preset and saved the video file at 'Super HQ' and the upscales turned out fine, but those files were already pretty sharp and Topaz had plenty to work with for increasing the resolution. I ran THE DEAD ZONE files through this same preset and the files definitely lost some grain as well, but again, the underlying image was good enough to upscale.
I wonder if my deinterlacing has been blurring the image and removing film grain from the Season 1 episodes. For other video sources, it was fine. But Season 1 was already blurry and lacking film grain; it couldn't have afforded to lose anything. I'm going to re-copy the disc files again and review the raw files.
Interesting. And, not that it's a problem, but you have to wonder if there was any undesired loss on s2 or even s3 and to what degree. S2's video quality on the dvds still look more like s1 than s4 or s5. S3 feels like a turning point and as you said elsewhere, s4&5 (the cheapest production) is what would have been the ideal scenario across all season's from a video quality standpoint.
I'd actually like to try another experiment.
Originally, I was only upscaling SLIDERS episodes to 720p and got okay results with Season 1 and great results with Season 2 onward. Season 1 episodes looked decent in mediums and closeups but like watercolour paintings in the wide shots.
After that, I read a bunch of articles from the DOCTOR WHO restoration workers upscaling old 1970s DOCTOR WHO episodes from 650 line analog videotape to 1080p for blu-ray releases. Their preference was to aim for 1080p so that the upscaling would be controlled in terms of their work on the video file and not by the television, so I began outputting my SLIDERS upscales to 1080p.
With episodes 102 - 109 of SLIDERS, however, I think that's proven to be a mistake because SLIDERS analog videotape episodes are 250 line videotape, 38.5 per cent the resolution of a 1970s DOCTOR WHO episodes. I haven't been happy with the AI upscales, but that's probably because I've been stretching the 250 line videotape files too far. These are 480p files; increasing them by 125 per cent via AI is too much whereas increasing them by a modest 50 per cent via AI is more achievable.
Originally, I upscaled the episodes from 480p to 720p using Topaz Artemis LQ for maximum compression artifact removal and sharpening, but I think that was also too far; there isn't enough to sharpen and it looks like a watercolour in wide shots. Topaz Artemis HQ doesn't try to sharpen too much; it's assuming that underneath the compression, the video won't benefit from further sharpening which is correct for SLIDERS' episodes 102 - 109 because they're too blurry for sharpening.
One of my subsequent experiments was running 102 - 109 through Topaz but leaving it a 480p file that had been deartifacted and denoised and then Lanczos scaling this version to 1080p. The results were good but inconsistent: wide shots looked fragmented or blurry while mediums and closeups looked decent.
"One of my subsequent experiments was running 102 - 109 through Topaz but leaving it a 480p file that had been deartifacted and denoised and then Lanczos scaling this version to 1080p."
This means without artemis, correct? eg Topaz just deartifacted and denoised, correct? Or was that done via artemis HQ?
When running the raw DVD files through Lanczos without Topaz, the 1080p versions were not as sharp in mediums and closeups and had compression noise, but the video quality was consistent across all shots instead of going from clear to blurry. The noise actually helped as a sort of filler data cross the blurry shots.
I'm wondering if using Topaz AI's Artemis HQ to remove the compression raise the resolution by 50 per cent (without trying to sharpen what isn't there like the Artemis LQ version) will yield at least a modest improvement on the initial 480p to 720p upscale.
So you mean running topaz HQ *after* upscaling via lanczos? (edit: actually i think that's your normal workflow)
I'm wondering if then using bicubic scaling to raise the resolution by another 50 per cent to 1080p will also be effective.
It will definitely add a little pixelation back into the image, but that will hopefully make all the wides, mediums and closeups look more consistent, with the pixelation serving a similar function to the compression noise.
I'm going to try it with "Luck of the Draw" and "Prince of Wails" which, to me, had the worst looking wide shots after AI upscaling.
Interesting... so are you suggesting lanczos followed by artemis hq followed by bicubic scaling?
The observation you made about not going to 1080 on s1-9 is interestiing. The AI just can't work with the content as well.
Just as a note in general, I observed a pretty intriguing thing the other day. The Jodorowsky's Dune BOOK / bible that was made when he was trying to pitch hollywood a feature in the 70s was bought for $3m by a bunch of cryto/nft bros recently. There are a limit number of books out there. They then and went raised 12 MILLION dollars from the nft/crypto community to basically share ownership of a future project to make tv/film content based on the book (inspired). Yes, without dune rights etc. Not using the brand I guess. The "investors" don't even get a share of profit as I understand it, and it's supposedly not profit driven or whatever.
They are even looking at costs of an animated series -- supposedly most series for 13 episodes cost $3m to make though I am sure it at least doubles at times.
It's all very interesting to me because I think about our own community. Like, I don't think we can repeat any of that, but would we be able to raise say 100-150k for a true restoration of S1? If our show was cooler to 20-30 year olds with expendable income and overlap into the nft/crypto world, I'd say yea... sure. But our audience is generally at least 40 and 35 youngest, and don't have the same expendable income or risk tolerance.
MY thing is, at the end of the day, a true scan of the negative of s1, even if you didn't put in the special effects, i think would be eye-opening to the community and it would feel newish. Maybe you could even go 16:9 if cameras and boom mikes are not in the way. It's the sorta passion project I wish we could rally around. Because I don't think it will happen by companies motivated by profit.
That said, I also think an alternative is a fan related project that would essentially do it in an upscaling way. If scans couldn't be made but let's say, we raised enough money to actually get some license to pay uni (my guess is it would minmally be $50k -- maybe just 25k for s1, i dont know) and then released the title as an actual blu-ray and you hope to break even.
In that case, I would think an approach of treating wide shots with its own up scale and medium / close with their own and compiling it on a timeline would be needed for the best look for s1.
All this is fantasy of course, but you just wonder what we as a community could pull together if we say could get a little push from a jerry or john. I would also very much like to see distributors, like Peacock, asking Universal to at minimum do an upscale job (though clearly it would not meet the quality of yours unless they somehow had access to better raw assets).
Tuesday at worst, tomorrow night at best.
