Sliders DVD Releases (Universal, Mill Creek, SD blu-ray, Restoration) (Page 15 of 15)

After the #SlidersRewatch last night, I keep noticing the quality of the Roku video keeps getting worse. It's like they continually run Sliders off of VHS tapes.

I'm still having trouble getting my Android TV MiBox to play the 60fps files made with pneumatic's script, even when reducing the bitrate to 4mbps. The files would play for a few minutes, but then freeze up and the player would crash. I don't think it's anything pneumatic did or didn't do because the files play fine on my laptops (an i7 processor with 32GB of RAM and an Nvidia GPU, an i5 processor with 16GB of RAM). I think the Android TV MiBox just can't handle a 60fps MP4 at a 4mbps bitrate.

I'm still keeping the original 8mbps files, but I'm currently experimenting with running two sets of encodes: one is 1080p 60fps at 2.5mbbps. However, I am seeing that the sharpening and the compression are making consistent areas of one colour (like sky) rather blocky. It might be irrelevant when viewed on a TV at living room distance.

Another option might be to output the file at a higher bitrate, maybe back at the original 8mbps recommended by pneumatic, but at 720p with the AndroidTV scaling the file back to 1080p. The lower resolution may make no difference; the original files looked more like 360i than 480i.

I shouldn't have done it in 60fps in the first place - that was just to preserve the rare occasional 60fps elements as I'm a stickler for that. 

After you reported the issue with 60fps not playing on your device, I rendered another one at 24fps here (script included): https://sliders.tv/bboard/viewtopic.php … 463#p14463

The "# 24fps IVTC" comment is the relevant bit.  The only difference with the 60fps version was to do a DoubleWeave() before the TFM(), everything else was the same.

I read about 3:2 pulldown: https://en.wikipedia.org/wiki/Three-two_pull_down

I understand now that SLIDERS would have been shot on film but converted to 29.98fps via 3:2 pulldown for broadcast television and 24fps was the native framerate of the original film image.

I think this means I should redo the Pilot at 24fps.

Sliders is on a streaming app called Xumo now as Universal made the title part of its library for FAST outlets.  Previously the title was made for exclusive (it ended up on roku channel only bc of peacocks deal to be allowed on roku platform itself.

The good news is the image quality of the material on Xumo looks better at first glance to my eye.  Either xumo or Universal did a better job with ingestion or distribution, respectfully.   It just does not have all the blurriness or artifacts or crazy whatever that is on the roku channel version for example

I did notice that the show felt like it had a more video like quality however.  I've seen this before on some TV modes where they try to insert in extra frames or blur the motion between frames.  And you get something that feels like it's less so shot on film and more so on handheld video camera.

It definitely looks better than Roku. Probably better than Peacock too but I don't use it anymore. Xumo looks more saturated which I don't mind either way. I compared it with the original DVD Universal release, and the DVD looks like blu-ray in comparison. So far it appears Xumo is the best streaming between the three.

I could tell with the Pilot and other episodes it was like night and day.

The Pilot is not the best episode for comparing video quality across distributors and streamers. It was edited on film in contrast to the rest of Season 1 being edited on analog videotape and Seasons 2 - 5 being edited on digital videotape.

**

Streaming Compression

Disclaimer 1: My understanding of all this is a little shaky, so please be aware that I may have made many errors as I try to explain a confusing situation to myself.

I haven't watched SLIDERS on streaming, but every streaming service uses video compression. I've read that Netflix compresses its video into a lossy format that loses some sharpness and eliminates most of the film grain to reduce bandwidth. They transmit the video. A sharpening and noise-addition layer is applied to the received stream. They are probably doing this for all content and not selectively. This approach is fine for HD digital video or film converted to HD digital video.

However, if the video is standard definition, Netflix's approach is going to diminish whatever limited merits the SD file had in the first place: what was already blurry is blurrier and sharpening makes it blockier.

Standardized Streaming

There's also Netflix's standardized streaming format: 24 fps (actually 23.976)with videos in a 1:1 pixel aspect. On average, DVDs of older TV shows would be 30 fps (well, 29.97), in 720x480 resolution in a 9:8 pixel ratio (I'm basing that on what I've read about DEEP SPACE NINE DVDs, but pneumatic can correct me).

Netflix's pixel aspect conversion turns 720x480 files into 576x432. Then Netflix applies inverse telecine to get 30 fps frame rates to back to 24 fps. The reduction creates blur, and the inverse telecine makes things worse for SD TV shows specifically.

3:2 Pulldown

From what I've read of pneumatic's work and read on my own: a lot of film-based projects were converted for home release, changed from 24 fps to 30 fps to match the 30 fps of a CRT television.

When playing meant-for-CRT 30 fps files on modern HDTV displays, the interlaced 30 fps video will stutter without conversion. Inverse telecine converts 30 fps video back into 24 fps. Inverse telecine combines multiple frames into single frames to restore the original frame rate. This is effective for playing 30 fps content that was originally shot and edited in 24 fps film.

However, most 90s TV shows were only shot on 24 fps film, not edited on 24 fps film. Instead, the 24 fps film was converted to 30 fps videotape via 3:2 pulldown. The conversion created new frames: three single frames, one duplicated frame (3:2), to reach 30 fps, and also separated the frames into interlaced fields for interlaced CRT display.

Jumble

As long as there's a consistent cadence of three single frames and one duplicated frame, the video looks consistent. Unfortunately, production would edit the finished episode together by copying different shots from different videotapes onto the video master. In addition, the master videotape would have sequences that only ever existed in a 30 fps format: special effects, specific video effects sequences.

I'm guessing here, but it looks to me like each transfer from analog tape to analog tape seemed to distort the cadence further as fields and frames were mis-duplicated or dropped, further confusing the cadence.

The final episode would be a jumble of inconsistent cadences that might look fine on a CRT TV but shows artifacts and mismatches when a player or streaming services uses inverse telecine. The confused cadence means that a streamer or a disc player using inverse telecine doesn't have a consistent cadence for combining mismatched frames for 24 fps display.

Image Loss

As a result: there's resolution loss when half of the fields are absent. There are artifacts when mismatched frames are combined. This is probably why pneumatic kept seeing interlacing issues that he said looked "baked in" and why the frame rates and cadence were so inconsistent within individual files. And because streaming is compressing the file as well, it looks even worse than DVD.

Telecine Equipment

The jagged edges are periodic in "Summer of Love", "Last Days", "The Weaker Sex" and "The King is Back" but severe in "Prince of Wails", "Fever", "Eggheads" and "Last Days". My guess is different episodes used different telecine hardware for 3:2 pulldown conversion, some of which created the 30 fps frame rate with poor frame duplication and inconsistent cadence, and some of which did a better job.

"Last Days", for example, suffers from telecine judder where the others don't, so it was transferred on different equipment than the rest. There are different telecine tools: flying spot scanners, line array CCDs, pulsed LED systems. Give the inconsistency, it's likely that Season 1 of SLIDERS used whatever telecine tools were available at the exact moment that film came in, and not every episode used the same process.

I think it is also likely that whatever process was used to transfer the Season 1 episodes' analog videotapes into a digital format for DVD release created another level of generational cadence distortion.

Season 2 Upgrades

Most of these issues seem to vanish with "Into the Mystic". Season 2 benefitted from switching to digital videotape and with that would have come a new film-to-videotape telecine process.

I'd be curious to know how SLIDERS in Seasons 2 - 5 look on a streaming service. With the DVDs, I don't recall any of the Season 1 interlacing issues on Seasons 2 - 5. They probably exist, but they don't seem as noticeable.

Two possibilities present themselves: digital videotape has some final process to make the cadences more consistent. Another possibility is that the telecine process for digital videotape ensured that sequences could be transferred from editing videotapes to the video master to DVD digital files with a lower level of generational loss or cadence distortion. This would isolate cadence issues to sequences where material from different videotapes are put in sequence with each other.

It could be some combination of both.

Possible Solutions

I have to think that Netflix's approach is likely the industry standard and other streamers would just compress more or compress less.

Probably, the solution could be at the studio level: studios can use a more pneumatic-style approach to their old SD 30 fps video library. They could get their material back to 24 fps themselves via QTGMC & TIVTC which provide field weaving, frame matching and selective frame discarding rather than inverse telecine assuming the cadence is consistent. They could use neural network upscaling to bring it to 1080p in a 1:1 pixel aspect ratio. They could provide those files to streamers and streamers won't do downscaling or frame rate conversion.

Disclaimer 2: Please please please please please do not ask me to contact NBCUniversal about this. It's not my job to provide unpaid labour to a multinational corporation's video on demand department. I care about an intellectual property that happens to exist on one of their ledgers, but that doesn't make me their slave.

I just watched a season 5 episode and actually was surprised.  Though that season has always looked good on DVD whereas s2 not so much, on xoom, the episode I watched (the one on the aircraft carrier or battleship) was closer to s2 quality on the same service than I expected.  Not 1 to 1 but still gap wasn't as significant as dvds.

I must say, looking back at your S2 up-res samples, season 2 really was quite an impressive leap (looking like blu-ray compared to what is on streaming).  And definitely on a higher level than s1.

It would be great if eventually with the advent of future technology, A.I. will be able to build out the rest of a 4:3 frame to go to 16:9.   It seems plausible this will eventually happen and I wonder if software companies like Topaz will see enough of a market/demand to introduce a feature like that. 

Right now though many of the seasons are resilient enough to withstand zooming/cropping to 16:9.  The image is clean enough, even if the cinematography gets affected a bit.

I'd be so disappointed if I saw that on one of my favorite shows. I still feel a show should only be presented in 4:3 if they find problems on the left and right sides of the film, rather than cropping it down. It would be a good idea if the studios allowed an option to toggle between a 4:3 and 16:9 version. The laziness of not paying attention and cropping actors heads is sad.

I'm glad you brought this up. This is something I've been contemplating with regards to SLIDERS.

I personally don't feel Topaz is really adding anything that isn't originally there. I don't think Jim_Hall feels the same way, and that's very reasonable.

Jim_Hall and I discussed upscaling as pertaining to facial enhancement on still images. That's a slightly different situation: we were using AI to sharpen up the faces on old SLIDERS publicity shots that Jim_Hall had bought and scanned and restored with painstaking effort and care.

Topaz Gigapixel rebuilds faces well, but they often look sharper than the rest of the photo. There was often a razor-sharp face on a blurrier neck with fuzzy hair and clothes. We found that for the most part, we had to erase most of the facial enhancements and kept only a small amount of sharpened facial elements, and it was often subtle to the point where we might as well have not done it. In one area where we differed: I preferred to keep the AI facially-upscaled eyes and mouths. Jim_Hall preferred to not even have that.

I deferred to Jim_Hall entirely on this as the photos were for his site and they were his photo scans.

But to me, Topaz is 'just' stretching the images, albeit in a highly intricate and detailed way that uses different methods for different image elements, and machine learning is clearly more reliable on rebuilding faces in photos.

Topaz uses a massive library of algorithmic functions with multiple function sets for specific textures (grass, skin, metal, concrete) and visual elements (eyes, hair) to add new pixels effectively. Ultimately, it's a way to stretch the image to a larger size. It isn't using the same technique for each part of the image, but multiple techniques on each element of the image. It isn't adding new content, just supporting existing content. Sometimes, it fails because Topaz is dependent on grain and SD video often lacks the grain needed for a quality AI upscale.

I don't feel Topaz AI as it exists would -- or even could -- change historical footage via upscaling. It would be stretching it, ableit in a complex and dynamic way.

But what if that changed?

We are at a point where AI can generate images based on text-based prompts. I think AI image generation could ultimately serve to recreate the videotape damaged episodes of SLIDERS, episodes 1.02 - 1.09 of Season 1. As pneumatic and I have noted: the videotape masters for those episodes don't even have an SD level of detail. AI sharpening doesn't work on them because the 240 line videotape format has eliminated all the film grain. Restoration is not possible. One future solution is using AI to engage in frame regeneration.

AI image generation will eventually be able to use images as prompts, and to look at a low resolution image and then generate a closely matching high res copy. One could convert episodes 1.02 - 1.09 into individual frames (24 frames per second) to produce a still image for each frame of the episodes. These stills could conceivably be fed into an AI algorithm to generate an HD image with the same elements and composition and colour, but with extrapolated sharpness and texture to replace what's missing within the low resolution image.

This would not be building on existing pixels to stretch the frame to a larger size like Topaz. Instead, it would be creating a new image that regenerates all the SD elements at HD resolution with AI taking some guesses to fill in an HD level of detail on clothing, hair, skin, props, textures, etc.. These replicated frames could then be reassembled into a video that matches the original audio. This would not be an upscale. This would be a reconstruction.

AI images right now can suffer from looking more like digital illustration than actual photography, but it's improving. Duplicating existing frame images would make it easier to generate photorealism. AI can already receive image input; eventually, it'll be able to generate a new version of the image input at a larger scale.

I wouldn't want studios to create HD versions of SD shows via AI reconstruction, but it might be good for special effects shots and missing film. One of the greatest difficulties in bringing older TV shows to HD: the special effects don't exist on film, but on SD videotape. STAR TREK and STAR TREK: THE NEXT GENERATION had to have all the effects rebuilt for HD. It proved so expensive that most studios today are not rebuilding SD special effects in HD, but upscaling the SD effects shots. Upscaled special effects shots tend to have a fuzziness and lack of crisp definition due to the lower resolution and blur of older generation video effects and computer graphics.

For THE X-FILES in HD, FOX was not able to find a number film reels. Special effects background plates, stock footage and backgrounds were often missing from the film, along with more mundane sequences. While the restoration team rebuilt the effects where they had the film to do so, in some cases, they had to resort to upscaled SD videotape that didn't match the rescanned film around it.

BABYLON 5 and LOIS AND CLARK: THE NEW ADVENTURES OF SUPERMAN, while rescanned and reassembled from film, didn't attempt to rebuilt any effects at all. Instead, they used upscaled SD effects shots to save money. This looks okay in BABYLON 5 because those original effects looked like Playstation 2 even when they first aired. But for LOIS AND CLARK, any time Superman speeds or flies or uses heat vision, it goes from blu-ray quality to VHS. I think those are areas where, in the future, missing film and special effects shots could use AI reconstructions generated from SD image inputs.

It would disappoint me if NBCUniversal went the AI upscale and reconstruction route for all episodes of SLIDERS. A studio should rescan the film and rebuild the episodes, and they should limit AI frame reconstruction to effects shots.

But on a fan level -- if NBCUniversal isn't going to remaster SLIDERS, then I think it would be a great fan project to use AI reconstruction on the low-res videotape episodes of Season 1.