I just came on her to mention something in the testing build with respect to VC-1
I’ve been watching some episode remuxes in VC-1 with chapters and at the chapter point the screen goes black and rebuffers before resuming normally.
My skin displays the chapters so it’s fairly obvious the connection to the chapters as the seek bar comes up.
I’d need to do some testing to eliminate add-on stuff and test for repeatability.
But it was happening pretty consistently and had me confused until I noticed the different codec.
I can post back with more info or links to tests files if needed.
Edit:
So I renamed my Kodi profile and tested this without any add-ons or anything else running and it played fine. No black screen/buffering at the chapters. It’s a sitcom so the chapters are at add breaks and cut to black anyway, so that I think that accounted for the black screen.
But yesterday it was definitely dropping out of the stream and was consistent with the chapter points
I put everything back and tried it again with the same VC-1 files with chapters and it’s also not doing it.
So I’m not sure what was going on yesterday. I’m pretty sure there is a bunch of tvdb stuff that’s now broken. So I’m sure that doesn’t help
But that shouldn’t really have caused any playback issues.
So I’m chalking it up to a random event and therefore I don’t think VC-1 files with chapters need any specific investigation.
Sure. I’ll stop asking about MPEG-2 stuff as well.
I’m obviously not communicating very well, here.
If the original material was captured as progressive (25p), then setting deinterlacing to Off fixes all deinterlacing artefacts. As previously discussed, as of right now, if you are watching 1080i/50 VC-1 and set deinterlacing to Off, this causes stutter (especially with output at 50Hz). But if you could fix the stutter problem, everything would then be perfect on stuff captured as 25p.
So, if what you’re asking is “Does setting deinterlacing to off fix all remaining issues with VC-1 stuff captured as 25p, apart from the stutter?” then yes, it does.
But what I think you’re actually asking is “Can we set deinterlacing to Off for all 1080i/50 or 1080i/60 VC-1 videos, and not even bother having the auto-select or deinterlace setting?” I guess I would have to say no to that. My guess is that most 1080i/50 VC-1 material was captured as 25p, and most 1080i/60 VC-1 was captured as 30p or 24p; but there’s bound to be something out there that wasn’t, and that will cause problems.
There’s a more common issue that some stuff, while filmed as 25p, has things like the text of the end credits generated in 50i. (There are several old BBC releases for which this happens). Speaking purely for myself, if the end credits are a little unclear but the episode/film itself is pristine, I’m absolutely fine with that; but I suspect somebody would raise it as a bug eventually.
Those disks were released quite recently, so they wouldn’t be VC-1.
But this, yes.
@grahamh I’ll try and capture something for you… might take me a few hours.
@grahamh and @Chillbo - here’s another extract from an episode of Little Dorrit. From roughly 00:00 to 00:36 is the end of the main episode, filmed as 25p. From there to the end is the titles, which were computer-generated as 50i text. So, if you set deintertlacing to Off and the output to 1080p/25Hz the live-action part will look perfect, but the text in the end credits will look a little odd - if you pause it during the credits then you’ll see a double image of each line of text.
With deinterlacing set to Deinterlace, the end-credit text should be perfect.
Also @grahamh and @Chillbo - while looking through a few things to find a useful extract, I noticed that my blu ray of The End of Time - the final David Tennant Doctor Who story - seems to be native 1080i/60 VC-1, i.e. captured as 60i, not 24p or 30p. It definitely produces artefacts with deinterlacing set to Off. I can give you an extract of that too, if you like…? (Weird that it should be 1080i/60 and not 1080i/50 - maybe I imported it?) Anyway, that’s a tangible example of something where you can’t set deinterlacing to Off.
Thanks for this and both of you for your patience. The whole nuts and bolts of how Auntie captures content and commits to disk was very well set out by @Steve_Neal about 3(?) years ago. The missing piece of the jigsaw is can we read each frame header and identify whether it should be ‘deinterlaced’ or not. There is more than one bit of metadata where that can be signalled, so the more clips we can get (preferably with mixed content) to analyse the better. It appears AML’s firmware fails to identify the type of field it’s dealing with but if we can put our finger on exactly which bit of metadata we can trust we can give it a helping hand, and ‘Auto’ should do what you would expect it to do.
BTW, when you say ‘it was captured at …’ are you just going on how it looks on the screen or do you have more evidence (eg from mediainfo?)
If I was a Dr Who fan with a BD collection this would be easier but that’s not the case and I don’t want to rely on any downloads which could have had anything done to them.
There is one David Tennant clip in our library at 29.9Hz involving a monster and a long rope.
Sorry, missed this. The video is definitely 60Hz and not 50. The fact that it’s native 60i and not 24p or 30p is something I’m judging by eye - if I turn deinterlacing off, there are visible combing artefacts.
As far as Who is concerned, I’ve got the specials (the four episodes in between seasons 4 and 5) plus seasons 5-8 on blu ray. If you want me to rip some more stuff (using makemkv) I’ve no objection, but I would imagine later Who episodes are quite likely coded the same way as The Eleventh Hour. I can also give you a bit more Little Dorrit if it’s useful. I should also have Cranford and Bleak House somewhere, though I might have to do some excavating(!) - they’re about the same vintage, so they’re probably VC-1 and may have some mixed content in the credits. The little extract I’ve given you before from Planet Earth is also 1080i VC-1 (although pure 25p, not mixed) - I don’t have the discs for that one any more, I’m afraid. (That one is a good test for stutter, incidentally, as it has a lot of very smooth camera movement).
Oh, not at all! I’m sorry to keep nagging you guys about it.
I would advise against use of the UK blu-rays of the David Tennant specials for deinterlacing testing. They are a botched standards conversion releasesd as a cost-saving measure. I don’t have the link to hand but I recall reading an analysis that showed how the field structure in these releases is an utter mess. I’ve recently picked up the German bluray set as that is 50i h264 (there are some burned-in German captions unfortunately but these are few and far between, and there is a vast improvement over the UK release, although “the Next Doctor” is DVD not bluray).
I posted that from “The Next Doctor” UK blu-ray. Iti’s an example of a VC-1 problem on the Vero I’ve only encountered with the David Tennant specials, where there is some occasional split-second partial corruption of the video. I’ve not seen this via other platforms, but the bluray in this case is a standards conversion so perhaps not the best reference to work with.
That’s interesting! It seemed unlikely that The End of Time would have been filmed at 60i, but at the same time the disc I’ve got is definitely 60Hz, and definitely gives combing artefacts with deinterlacing turned off. If it’s actually a 50->60Hz conversion, that would explain it.
That said, it’s still an example of where you can’t just turn deinterlacing off and expect everything to work.
It was shot in SD and upscaled for the blu ray, so you aren’t missing that much.
thanks, have shared a link to some mixed-content VC-1.
I go by production date and look. The difference in look between content captured/authored as progressive compared to natively interlaced is really substantial. The only interlaced content I can think of on bluray would be sit-coms or new up-scaled releases of old shows. Any drama show for some years now will be progressive even if it’s 50i on disc, I recall someone far more qualified posting on this a while back.
Here are a few more samples to look at (to go with the three from higher up the thread).
Bleak House and Cranford Season 1 seem to have gone through the same 50->60Hz conversion process as the Doctor Who specials, so the video needs to be played as 1080i/60 with deinterlacing turned on (“deinterlace”).
Cranford season 2 and the three bits from seasons 1-3 of Being Human are 25p in the epsiode transitioning to 50i text in the credits. In the Cranford extract there are about two seconds where the 50i text is playing over the top of a fading 25p background; in Being Human there’s a sharp transition.
Seasons 4 and 5 of Being Human are 1080i/50 h.264, and so is season 6 of Doctor Who (and presumably later seasons), so obviously not applicable for VC-1 testing. If you would like some more pieces of anything I’ve recently posted extracts of, let me know.
What we’re doing now is changing the ‘deinterlace’ method on the fly. So those BBC clips generally play in ‘progressive’ mode (officially ‘frame interlace’) but switch to ‘field interlace’ for the credits. You need to set the setting to Auto. The other two options override the automatic detection. You will need that with the 1080_60i_VC1 clip.
The check is done about every second and I’ve cut out some code that was running in Kodi on every frame in the hope it might improve stutter. Not sure it’s had an effect - vertical pans can still be a bit jerky, but this could be display-dependent (motion compensation?).
For the curious, you will see in the Kodi log when the interlace handling method changes.
And before you ask, to do something similar for h264 is waaay more complicated.
Hello, @grahamh . I’m afraid the stutter problem (with deinterlacing set to Off) is still visible. It’s rather intermittent - just now I played my Planet Earth test clip, and it was smooth three times in a row, but stuttered badly the fourth time. That particular clip always seems to play smoothly with 25Hz output, though. (And it’s only with deinterlacing set to Off that I’m seeing stutter).
Otherwise, things generally seem to look quite good with deinterlacing set to Auto-Select. I didn’t notice any combing on the 60Hz clips with that setting.
I did notice one slight artefact on my old Little Dorrit clip (not the one I uploaded a few days ago, but the one that’s available at the link two paragraphs ago in this post). At about 00:10-00:14 there’s a shot of the piano from the side, with a rectangular shape inlaid in gold. Above the gold rectangle (but below the music stand), there are four bright horizontal bands, where the light reflects on curves in the piano lid. If you watch those reflections carefully, then with deinterlacing set to Auto-Select there is some visible flicker/shimmer; with deinterlacing set to Off the shimmer disappears.
For completeness: I also spotted some artefacts in my clip from the Doctor Who story The End of Time. (The previous download link for that one has expired, but if you didn’t grab it before I can upload it again). Starting from about 00:56 there’s the title sequence with the TARDIS flying through a brightly coloured tunnel. There are a couple of places in there where I can see a very quick burst of image corruption - it looks a bit like macro-blocking. (On my Oppo blu ray player, I don’t see that).
Okay. I hope, though, that (after getting VC-1 nailed down) you guys will find the time to look at the other deinterlacing regressions that came in with the 4.9 kernel, especially the lack of diagonal filtering - given that this used to work, it seems like it ought to be possible to get it working again.
You might also like to check out the “1080i_60_deinterlace_test” from Dropbox - Deinterlacing test patterns - Simplify your life -that’s the wedge pattern you’ve seen before. That one also doesn’t play correctly with deinterlacing set to Auto-Select, but does play correctly with deinterlacing set to Off.
That’s … odd. As far as the kernel is concerned they are the same. Can you check
cat /sys/module/amvdec_vc1/parameters/force_frameint while playing with Auto and Off?
I just tried that clip on my 4k TV and can’t tell the difference between Auto and Off. With the display at 4k, there’s a little moire just above 1080 and I guess that’s inevitable. My TV also has a 1->4 pixel mode so you could say it’s a FHD screen. With that I get an extra moire pattern about half way between 1080 and 540. With either setting, Kodi recognises it’s a pull-down and switches to 24Hz after 7 seconds. If I de-whitelist 24Hz there’s a little glitch at 7 seconds as if Kodi’s thinking about changing the refreshrate. Then it’s fine.
Thanks for working on this. Auto looks to be doing a good job with mixed 50i VC-1, clean on progressive equivalent to disabled deinterlace and handles the transition to interlaced credits. Some stutter still, both on auto and deinterlace set to off.
I’ll stop asking about MPEG-2 stuff as well. 
Not perfect at the same time - we’d hoped for a different reply, obviously. 
