Moving from Infuse to MrMC: Missing a few things

[davilla]

av_probe_input_buffer

Just push the sample to https://nextcloud.semperpax.be/s/DeeCbJz3AMDFRD9

Thanks, I've uploaded F1 AZB GP Snip 1b.ts (92.0MB, 1 min).

Can I invoke av_probe_input_buffer from an ffmpeg command line?

mediainfo says:

Code: Select all

Video
ID                                       : 256 (0x100)
Menu ID                                  : 1 (0x1)
Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format profile                           : High@L4
Format settings                          : CABAC / 4 Ref Frames
Format settings, CABAC                   : Yes
Format settings, ReFrames                : 4 frames
Codec ID                                 : 27
Duration                                 : 59 s 480 ms
Width                                    : 1 920 pixels
Height                                   : 1 080 pixels
Display aspect ratio                     : 16:9
Frame rate mode                          : Variable
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8 bits
Scan type                                : Progressive

ffprobe says:
Stream #0:0[0x100]: Video: h264 (High) ([27][0][0][0] / 0x001B), yuv420p, 1920x1080 [SAR 1:1 DAR 16:9], 25 fps, 25 tbr, 90k tbn, 50 tbc

This video is not interlaced.

[MofTot]

av_probe_input_buffer

Just push the sample to https://nextcloud.semperpax.be/s/DeeCbJz3AMDFRD9

Thanks, I've uploaded F1 AZB GP Snip 1b.ts (92.0MB, 1 min).

Can I invoke av_probe_input_buffer from an ffmpeg command line?

mediainfo says:

Code: Select all

Video
ID                                       : 256 (0x100)
Menu ID                                  : 1 (0x1)
Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format profile                           : High@L4
Format settings                          : CABAC / 4 Ref Frames
Format settings, CABAC                   : Yes
Format settings, ReFrames                : 4 frames
Codec ID                                 : 27
Duration                                 : 59 s 480 ms
Width                                    : 1 920 pixels
Height                                   : 1 080 pixels
Display aspect ratio                     : 16:9
Frame rate mode                          : Variable
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8 bits
Scan type                                : Progressive

ffprobe says:
Stream #0:0[0x100]: Video: h264 (High) ([27][0][0][0] / 0x001B), yuv420p, 1920x1080 [SAR 1:1 DAR 16:9], 25 fps, 25 tbr, 90k tbn, 50 tbc

This video is not interlaced.

Sadly they're both wrong. Just look at it. Try it out as I suggested in VLC and you can see how the interlaced lines are removed when setting it to deinterlace. I'll upload two screenshots that demonstrate this (they're too large for attachment).

Uploaded:

F1 AZB GP Snip 1b 00;02;00 VLC Deinterlace Off.png
F1 AZB GP Snip 1b 00;02;00 VLC Deinterlace On.png

[Montoya]

Sadly they're both wrong.

FYI, in MediaInfo, my recordings of a 1080i TV channel are reported like this:

Code: Select all

Scan type                                : Interlaced
Scan type, store method                  : Separated fields
Scan order                               : Top Field First

So where is your "scan type store method/scan order", which should be part of interlaced content ?

Below full media info of 1080i test recording:

Code: Select all

Video
ID                                       : 2001 (0x7D1)
Menu ID                                  : 19401 (0x4BC9)
Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format profile                           : High@L4
Format settings                          : CABAC / 4 Ref Frames
Format settings, CABAC                   : Yes
Format settings, Reference frames        : 4 frames
Codec ID                                 : 27
Duration                                 : 13 s 960 ms
Bit rate                                 : 12.7 Mb/s
Width                                    : 1 920 pixels
Height                                   : 1 080 pixels
Display aspect ratio                     : 16:9
Frame rate                               : 25.000 FPS
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8 bits
Scan type                                : Interlaced
Scan type, store method                  : Separated fields
Scan order                               : Top Field First
Bits/(Pixel*Frame)                       : 0.245
Stream size                              : 21.1 MiB (81%)
Color range                              : Limited
Color primaries                          : BT.709
Transfer characteristics                 : BT.709
Matrix coefficients                      : BT.709

[MofTot]

So where is your "scan type store method/scan order", which should be part of interlaced content ?

The issue is that mediainfo only looks at the metadata, not the video frames. So if the metadata for some reason is incorrect (eg if an 1080i broadcast was "camouflaged" as 1080p) this will lead to the wrong conclusions.

Freeview UK Channel 4 HD doesn't broadcast F1 in 1080p25, sadly "only" in 1080i50, which is very unpleasant to watch unless it's deinterlaced.

Perhaps it would be worth allowing expert users to somehow override the auto detection of interlaced content. Or even smarter, allow expert users to force deinterlacing for all content in a list of given source folders.

[davilla]

There are two flavors of deinterlaced h264 content.

For the 1st, the entire content is flagged as deinterlaced. It's easy to detect at the probe phase of analyzing media before playback.

The 2nd type is much harder to detect as it can only be determined by checking the decoded frames. To make matters even more difficult, this type can switch from interlaced and progressive on frame boundaries. So you could have N frames of interlaced, then Y frames of progressive, the Z frames of interlaced. When MrMC sees media flagged this way, it should assume that it's interlaced and kick it off to ffmpeg for sw decoding. The hw decoders just cannot deal with all the various interlaced flavors.

I suspect this particular content is the second type and It's actually switching on the fly (interlaced/progressive). First one that I've seen to do this. Have to poke deeper

[davilla]

The issue is that mediainfo only looks at the metadata, not the video frames. So if the metadata for some reason is incorrect (eg if an 1080i broadcast was "camouflaged" as 1080p) this will lead to the wrong conclusions.

Mediainfo does look at decoded content for info analysis. The problem is that it does not look over the entire video length, only a small part.

[Montoya]

Perhaps it would be worth allowing expert users to somehow override the auto detection of interlaced content. Or even smarter, allow expert users to force deinterlacing for all content in a list of given source folders.

There still is a hole present in MrMC you could try (and maybe works, or not, for your content), before it's plugged by Davilla :

1. Find interlaced sample content that MrMC detects
2. Play that interlaced sample content in MrMC
3. During play go to OSD Settings >> video
4. Set "Deinterlace video" to "on"
5. Set your preferred "Deinterlace method"
6. Scroll down to "Set as default for all media"
7. stop playing your sample interlaced content
8. play your own interlaced content and check if deinterlacing takes place now

WARNING:
Using the above brakes MrMC video playback of non-interlaced content and MrMC settings from step 4 and 5 from above list, can ONLY be reverted when interlaced content (that MrMC can detect) is played, by reverting the setting in step 4 to "auto" and step 5 to "Auto select".

[MofTot]

There are two flavors of deinterlaced h264 content.

For the 1st, the entire content is flagged as deinterlaced. It's easy to detect at the probe phase of analyzing media before playback.

The 2nd type is much harder to detect as it can only be determined by checking the decoded frames. To make matters even more difficult, this type can switch from interlaced and progressive on frame boundaries. So you could have N frames of interlaced, then Y frames of progressive, the Z frames of interlaced. When MrMC sees media flagged this way, it should assume that it's interlaced and kick it off to ffmpeg for sw decoding. The hw decoders just cannot deal with all the various interlaced flavors.

I suspect this particular content is the second type and It's actually switching on the fly (interlaced/progressive). First one that I've seen to do this. Have to poke deeper

I think there's something wrong with the snippet I provided (and other small snippets I have), apologies. I tried to force deinterlacing as it was suggested and that broke playback. I suggest you stop looking into this for now and I'll get back to you after I've analyzed playback of this weekend's recordings of F1 from Japan (typhoon permitting).

[MofTot]

The issue is that mediainfo only looks at the metadata, not the video frames. So if the metadata for some reason is incorrect (eg if an 1080i broadcast was "camouflaged" as 1080p) this will lead to the wrong conclusions.

Mediainfo does look at decoded content for info analysis. The problem is that it does not look over the entire video length, only a small part.

Hmm, that's contrary to what other people write on the Internet, don't know who to believe .

[MofTot]

Perhaps it would be worth allowing expert users to somehow override the auto detection of interlaced content. Or even smarter, allow expert users to force deinterlacing for all content in a list of given source folders.

There still is a hole present in MrMC you could try (and maybe works, or not, for your content), before it's plugged by Davilla :

1. Find interlaced sample content that MrMC detects
2. Play that interlaced sample content in MrMC
3. During play go to OSD Settings >> video
4. Set "Deinterlace video" to "on"
5. Set your preferred "Deinterlace method"
6. Scroll down to "Set as default for all media"
7. stop playing your sample interlaced content
8. play your own interlaced content and check if deinterlacing takes place now

WARNING:
Using the above brakes MrMC video playback of non-interlaced content and MrMC settings from step 4 and 5 from above list, can ONLY be reverted when interlaced content (that MrMC can detect) is played, by reverting the setting in step 4 to "auto" and step 5 to "Auto select".

Thanks a lot for your suggestion, that broke playback of what I believe is interlaced but MrMC isn't having it. There might be something entirely wrong with the snippets I'm using for testing, I'll get back next week after having analyzed new F1 recordings this weekend, typhoon permitting.