U-Matic music video transfers better than YouTube? +Some DUB logistical Qs

[aramkolt]

Since getting into U-Matic, I've collected some U-Matic music videos that were probably used by networks such as MTV to play over the air before betacam was the more preferred format. Most of these are from 1995 to 2000-ish. I got them mainly to use as test tapes because they were commercially produced on I'm sure good working hardware at the time.

I'm fairly sure that I can get better transfers of these music videos than what are currently on YouTube posted on their the official channels. My guess is Youtube's compression/bitrate is better than it was 15+ years ago when most of these videos were originally uploaded as 480p 29.97fps.

The videos on YouTube at least in some cases seem to be zoomed in or have ditched some of the overscan area in all directions, almost as though they were encoded by an elgato video capture or something (which are known to do that). U-Matic doesn't really have head switching noise, so it doesn't make sense to me why they'd have intentionally done that.

See attached screenshot of the same frame. YouTube version on the left, and my very basic composite capture on the right from EVE6 - Open Road Song.



Deinterlacing for the screenshot was just done with handbrake since this isn't a final product, so I'm sure it would look better with some Avisynth mastery and QTGMC, but that'd get rid of any perceived blockiness with a high enough bitrate I'm sure.

It should be possible to get even better quality with appropriately working DUB hardware, but I think I'll need to either adjust my hardware or acquire some more for that to be viable. If anyone has service manuals for DPS-295, 275, or 210, please let me know! Or if anyone has tips for getting DUB devices to work better, I'd like to hear any ideas. Currently what they are doing is top of screen tearing which indicates the need for a line TBC which admittedly these machines do not have, though since most of the DUB devices already are TBCs, putting a line TBC after a frame TBC isn't ideal.

Also of note, I'm not really seeing much ringing via composite which I've heard is usually pretty prevalent with U-Matic. My guess is the ringing will be more visible with DUB if I can get that going?

For my purposes, I can't see dropping $500 GBP for a keystroke dub optimser, but I'm aware it exists. If anyone has one, I'd appreciate seeing a picture of what's inside, my guess is that they are relatively simple, but I'm definitely not an electrical engineer haha. It's kind of interesting that they don't recommend using it for the BVU-950 right on their site, so that kind of suggests that certain machines might have composite output that is better than dub - maybe. I'm not sure why they didn't say that they don't recommend using it with any of the BVU-9x0 series though since they all had optional internal TBCs, though interestingly, if DUB output is used on them (including the BVU-950 from owners I've spoken to), the internal TBC gets bypassed completely.

If I am able to get better transfers and then post them to YouTube, I assume they'll just get immediately pulled for copyright? Or does YouTube allow them to stay up but just contributes any view-monetization towards the original copyright holder?

[dellsam34]

Most of those transfers on youtube also look blurrier than even a decent VHS transfer because of the stupid 720p conversion, I've seen some Vevo 480p that looked good, But off course now YT makes 480 and 720 uploads intentionally look like crap to save bandwidth and storage.

What uMatic machine do you have? I think Colin from video99 YT channel modded some of his machines, check out those videos.

[lordsmurf]

If I am able to get better transfers and then post them to YouTube, I assume they'll just get immediately pulled for copyright? Or does YouTube allow them to stay up but just contributes any view-monetization towards the original copyright holder?

That happens --- but sometimes it's just a strike against you, video taken down.

It's random. Youtube is chaos and randomness at times.

[Alwyn]

Or does YouTube allow them to stay up but just contributes any view-monetization towards the original copyright holder?

I have multiple YT channels for "testing". Put the music up on one of them, if you get a strike you'll know. If it gets a Copyright flag, it's all good, the owner makes money from you hosting their music. You obviously won't be able to. Then upload it to your main channel.

Re those screenshots, there's not a lot in it. The YT is softer. Whether it warrants a new upload...

[aramkolt]

Machines I have access to for testing are a VO-5800, VO9800, VP9000, and BVU900. The VP9000 doesn't have DUB output, but I have a feeling S-Video modding it (if successful) would be better than DUB and more versatile at least as you'd have a huge variety of TBCs to pair an S-Video output with compared to DUB.

I did give uploading a music video a test to see what YouTube would do. Basically it was flagged it for copyright and wouldn't allow the video to be posted at all. Since the video was never posted, it doesn't count as an official strike.

I'm definitely NOT wanting to make money on the videos themselves. Goal of posting a better transfer is to put it out there for historical preservation (and I'd prefer that any potential revenue goes to the copyright holder) AND to promote the quality of my U-matic transfers (which is technically fair use).

There's tons of U-Matic transfer services out there, but most don't really post capture comparisons as to why you should choose them over another option. It's mostly whatever jargon can be thrown into a description to make their transfers "sound" like they must be good. I would rather "show" than "tell". Even then, it's not really about making money, it's about justifying the purchases of a variety of hardware used and for the enjoyment of analog media preservation. Ideally it'd be nice just to break even on some of the hardware used - which will definitely take a while haha.

To demo the transfer quality, I don't actually need to post the whole video though and It'd probably need to be shorter 5 second clips of probably a variety of music videos from different tapes. Just would be nice to achieve both goals at once. If I was allowed to post full videos, I'd probably put a video of the original tape loading into the U-Matic Machine (cause that looks cool to see 50 year old technology at work), then the video capture, then a single slide at the very end of the video and a link in the description on how to contact me if a viewer wants to inquire about video transfer services.

Also, for some reason, I didn't realize that the Domesday duplicator also supports U-Matic, so my goal is also to get that working (currently having install issues on my Mac, but I'm sure Windows is easier to install and I may just cave and do that instead) and compare these 4 transfer results:

1. Composite output
2. Dub output
3. S-Video modded output (haven't achieved this yet, but I know that it is technically possible)
4. DomesdayDuplcator Capture

The first three can be captured all with the same capture card, though Dub output will need to go through a different device first to get it into either S-Video or Component Video that the capture card can accept.

Either way, seems lame there's no option to just direct revenue to copyright holders and still post videos (without copyright strikes). Only reason I can see for that not being an option is that YouTube would have to host a lot more "duplicate" videos that are really the same content more or less.

As far as the screenshots go, theoretically DUB, S-Video, and Domesday duplicator should all end up looking better than the preliminary composite result, though they might possibly have more ringing.

To get DUB working appropriately, I'll likely need some service manuals on how to adjust the DPS units, but they seem impossible to find for the 210, 275, and 295. The 295 is interesting because it derives chroma from composite which would seem to defeat the purpose of using DUB to begin with..... unless there's some advantage to taking a separate luma and composite signal and then being able to "subtract" the two, leaving a chroma signal that is as good as having never combined them, but I somehow doubt that is the case.

[Alwyn]

To demo the transfer quality, I don't actually need to post the whole video though and It'd probably need to be shorter 5 second clips of probably a variety of music videos from different tapes.

You could try YT "Shorts". I don't know but I think less than 30 seconds avoids the Copyright issue. For one sound track, I came across a list of the Shorts videos using it. There were dozens.

[aramkolt]

Got a bit further improvement out of the VP9000, doubt I'll be able to improve it more. Eventually I'll see what the DomesdayDuplicator can do. Youtube is on the left, my VP9000 is on the right. Technically this was done in PsF interlacing, so it's interlaced, but there's only motion every 2 fields, so the above screenshot was not deinterlaced and that's how the frame looks like coming off the player. Obviously I also didn't correct the pixel aspect ratio either, but if I did, should be some extra perceived horizontal resolution to the U-Matic capture also.

I actually reached out to Eve6 via email to see if the company that made the original video is still around (which would have the digital source files), If they aren't, or they no longer have the original files, this could be one of the better copies still left out there, so I asked if they wanted it to re-upload themselves.

[dellsam34]

That's a major improvement, If you see noise grains and chroma blotches that's called transparency, meaning your capture chain is not processing much of it, just passing it along with minimum degradation, The one on the left is what it would look like after post processing, which I'm not a big fan of.

[pandy]

Obviously it is heavily denoised - perhaps YT has ability to detect lowres but noisy video to apply heavy denoise to improve compressability - bandwidth and storage space are costly. Upscale video to HD or UHD and then try challenge YT.

[Alwyn]

I think a bit of Camcorder Colour Denoiser would help (shirt colour doesn't look right). Also, the AR doesn't look right. But definitely better than the old YT version. Looks like his face has had a tough life!

[timtape]

Most of those transfers on youtube also look blurrier than even a decent VHS transfer because of the stupid 720p conversion, I've seen some Vevo 480p that looked good, But off course now YT makes 480 and 720 uploads intentionally look like crap to save bandwidth and storage...

Yes YT at 480p does significantly degrade the picture quality, even a good VHS transfer. Audio seems not nearly so affected. I only really discovered how degraded after directly comparing the YT results with my files before uploading.

It's much more common for people these days to upload such standard def source material at 1080p. Anything I upload in future will be at a rate high enough that the YT compression has little or no effect on the picture quality.

[Alwyn]

But off course now YT makes 480 and 720 uploads intentionally look like crap to save bandwidth and storage.

Which is totally illogical because people just go and upload at much higher resolutions and bitrate to get around that "problem". If YT was trying to save anything, you'd think they'd be encouraging low bitrates, not punishing them.

[timtape]

I'm not sure on this because I haven't tested it. Is it possible YT is degrading every upload level by the same amount, let's say by 20%. Their 1080p may look great on its own but compared to what? It might degrade the 1080p upload the same % as the smaller ones, but without seeing the 1080p user upload before YT compression how can we be sure?

[dellsam34]

I'm not sure on this because I haven't tested it. Is it possible YT is degrading every upload level by the same amount, let's say by 20%. Their 1080p may look great on its own but compared to what? It might degrade the 1080p upload the same % as the smaller ones, but without seeing the 1080p user upload before YT compression how can we be sure?

It is unclear how YT algorith works, but they are prioritizing HD and 4k contents and putting anything less on the back burner as it is not important and just takes up space, So for now upscaling can fool the algorithm but not sure how that is going to last, I believe they will come up with some AI option to detect upscaled contents.

I've previously uploaded 3 video samples into YT with pure static to stress test the algorithm, 480p, 720p and 1080p, Download them and make your own analysis, I no longer have the original file though, so will just compare the 3 YT versions.