Media File playback misses the start of the file

Gully · Post by **Gully** » Wed May 11, 2016 16:59

Just wondering if anyone is finding this.

Still trying to narrow down the details but to use the last example of a MKV file (video is Advanced Video Coding, audio is AAC) when I start playing i, it starts about 5-6 seconds in. Only obvious when it starts mid sentence or movement. Nothing on the timeline, etc that would make you realise.

But playing on the PC and I can see what was missed.

I have tried manual and auto resolution just in case, but there is no sense or glimmer of missing footage.

dRdoS7 · Post by **dRdoS7** » Wed May 11, 2016 19:07

Just fire up the Tardis and go back to the start.

Star6key · Post by **Star6key** » Wed May 11, 2016 19:17

I only see that sort of behaviour if I'm editing a file (on the pc) and I don't cut on an I-frame. On playback, there will be a black screen until it gets to the first I-frame, then all is ok. This has nothing to do with the T4, but it sounds similar.

prl · Post by **prl** » Wed May 11, 2016 19:37

I-frames 5-6 saconds apart sounds a bit long. Also, that would affect any playback.

Gully · Post by **Gully** » Wed May 11, 2016 19:44

dRdoS7 wrote:Just fire up the Tardis and go back to the start.

Waste of energy and time. The start is lost somewhere in the dark realms of the Beyonwiz, not in time.

Star6key wrote:I only see that sort of behaviour if I'm editing a file (on the pc) and I don't cut on an I-frame. On playback, there will be a black screen until it gets to the first I-frame, then all is ok. This has nothing to do with the T4, but it sounds similar.

No black screen, just starts and misses a bit. And as Peter posted it would apply to all playback.

Gully · Post by **Gully** » Thu May 12, 2016 15:19

Testing more files and does seem to be limited to mp4 files, but consistent results of missing the start.

All mine are encoded in similar format, so not sure if that is a factor.

Has anyone else tested this?

Gully · Post by **Gully** » Thu May 12, 2016 18:50

Further testing and I have found some mp4 files that don't have the problem.

Looking at the files with MediaInfo, aside from some different bitrates, the main differences are

okay vs problem files

Bit rate mode variable vs constant
Format profile High@L3.1 vs Main@L3.1
Format settings, ReFrames 5 frames vs 2 frames
Frame rate 59.940 (60000/1001) fps vs 25.000 fps

Can anyone with better knowledge of video formats see anything here that gives a clue to the problem?

Happy to post more info if it helps.

bonson · Post by **bonson** » Thu May 12, 2016 22:56

Don't know much about video formats but I have found that the built-in media player in my Panasonic tv has a problem playing files with a Reframe rate less than 4. The file plays but the video playback stutters badly.

Edit: Just tried one of these "problem" files on my new T2 and it plays fine with no skipping at the start so you can ignore this whole post.

Gully · Post by **Gully** » Sat Jun 04, 2016 11:41

So the problem persists for me and I am wondering if there is anyone else experiencing the same problem.

Only mp4 files, not all of them, missing about 6 seconds at the start when I compare playback to my PC.

Guessing it is something in the encoding but I haven't isolated what.

Gully · Post by **Gully** » Mon Jun 13, 2016 16:19

Decided to do some testing to see when the firmware changed to introduce this problem.

First put the media file on a USB (same one as the firmware) and checked it and confirmed the problem.

Just reinstalled the previous public release 04/11/15 and the problem is not there any longer.

So I'll try a few more and see if I can isolate where it was introduced more precisely.

EDIT: Just tried 1/1/16 beta release and still problem free.

EDIT2: Tried the 5/4/16 beta and the problem is there so somewhere in between. I'll try the 25/1 beta just to narrow the timeline further.

EDIT3: The 25/1 beta is fine and problem free, so somewhere between there and the 5/4/16 version.

Any ideas/thoughts on what changes and how it can be reversed?

MrQuade · Post by **MrQuade** » Mon Jun 13, 2016 18:05

The most likely culprit would have been the changes to fix the lipsync issues we were seeing in the previous firmware revisions.
Prior to that, you would get large discrepancies between audio and video starting up, so I would say that there is something in that particular file that triggers the media player (with lipsync fixes) to start a few seconds into the video.

viewtopic.php?f=56&t=10048&start=20#p135056

Gully · Post by **Gully** » Mon Jun 13, 2016 22:01

Yes, I was wondering about that but didn't know enough about it to suggest it.

Or to exclude any other possible changes.

So hopefully if it is that, PeterU should know what the change was and if it is a likely cause. And even better a fix for it.

I'll add a post to the bug reporting section now as well then.

peteru · Post by **peteru** » Tue Jun 14, 2016 02:01

Can you dig out the driver versions for the working and non-working cases? Can we pin point it to a specific driver version?

I'm trying to establish whether the regression is due to an SDK change, a driver change or an application change.

Gully · Post by **Gully** » Tue Jun 14, 2016 10:55

Is there an easy way to do that without re-flashing the 2 firmwares?

I was just using the 2 USB releases without any online updates. (25/1 and 5/4)

MrQuade · Post by **MrQuade** » Tue Jun 14, 2016 16:15

peteru wrote:Can you dig out the driver versions for the working and non-working cases? Can we pin point it to a specific driver version?

I'm trying to establish whether the regression is due to an SDK change, a driver change or an application change.

I'm happy to do the firmware flashing as I have an afternoon to myself today.

What information do you need in terms of driver versions?
The Info->About just lists a date, is that all?

Gully · Post by **Gully** » Tue Jun 14, 2016 16:28

MrQuade wrote:I'm happy to do the firmware flashing as I have an afternoon to myself today.

Thanks MrQuade.

MrQuade · Post by **MrQuade** » Tue Jun 14, 2016 17:40

I've been running the 25/1/2016 firmware for a bit. I can definitely confirm that the lipsync is a bit dodgy in this version. I am just catching up on some old DP recordings that were converted to h.264 mkv.

Still not sure where to find the driver version, but all my timers have been set for the evening, so I can stay on this version for a while if required.

I can check on the T2/T4 too if needed, bit I don't have a confirmed test file to check if the issue is present on them anyway.

MrQuade · Post by **MrQuade** » Tue Jun 14, 2016 17:59

In the meantime, I can offer this if it helps:

Code: Select all

root@beyonwizt3:~# opkg list-installed *driver*
enigma2-plugin-drivers-dvb-usb-af9015 - 1.0-r6.73
enigma2-plugin-drivers-dvb-usb-af9035 - 1.0-r4.73
enigma2-plugin-drivers-dvb-usb-dib0700 - 1.0-r8.73
enigma2-plugin-drivers-dvb-usb-dtt200u - 1.0-r1.73
enigma2-plugin-drivers-dvb-usb-dw2102 - 1.0-r2.73
enigma2-plugin-drivers-dvb-usb-em28xx - 1.0-r4.73
enigma2-plugin-drivers-dvb-usb-it913x - 1.0-r2.73
enigma2-plugin-drivers-dvb-usb-pctv452e - 1.0-r1.73
enigma2-plugin-drivers-network-usb-ath9k-htc - 1.0-r2.73
enigma2-plugin-drivers-network-usb-carl9170 - 1.0-r2.73
enigma2-plugin-drivers-network-usb-r8712u - 1.0-r3.70
enigma2-plugin-drivers-network-usb-rt2500 - 1.0-r2.74
enigma2-plugin-drivers-network-usb-rt2800 - 1.0-r2.73
enigma2-plugin-drivers-network-usb-rt73 - 1.0-r3.70
enigma2-plugin-drivers-network-usb-rtl8187 - 1.0-r2.73
enigma2-plugin-drivers-network-usb-rtl8192cu - 1.0-r3.73
enigma2-plugin-drivers-network-usb-zd1211rw - 1.0-r2.73
enigma2-plugin-drivers-usbserial - 1.0-r0.70
oe-alliance-drivers - 1.0-r11.124

Gully · Post by **Gully** » Tue Jun 14, 2016 18:01

In terms of the problem, it hasn't been one I have seen with mkv containers which I find are fine with the current firmware, just mp4. (And not 100% of them but a majority of the files I have tried miss the start.)

peteru · Post by **peteru** » Tue Jun 14, 2016 21:11

MrQuade wrote:What information do you need in terms of driver versions?
The Info->About just lists a date, is that all?

That's what I need - the version / date string for the last known good version and then the first broken version.

MrQuade · Post by **MrQuade** » Tue Jun 14, 2016 21:50

peteru wrote: That's what I need - the version / date string for the last known good version and then the first broken version.

No worries.
The 25/1/2016 image was the last known working image, and the first broken one was 05/04/2016 according to Gully's post.

T3 firmware version http://feeds.beyonwiz.org/beyonwiz-4.4- ... 25_usb.zip
Drivers: 2016-01-25

T3 firmware version http://feeds.beyonwiz.org/beyonwiz-4.4- ... 05_usb.zip
Drivers: 2016-01-25

Looks like they are the same if they are being reported correctly. Is there anything else I can check?

Gully · Post by **Gully** » Tue Jun 14, 2016 22:15

Sounds like it isn't the drivers if that is correct then.

peteru · Post by **peteru** » Tue Jun 14, 2016 23:35

Is this a T3 only problem?

MrQuade · Post by **MrQuade** » Tue Jun 14, 2016 23:43

I am happy to check, but I can't find any .mp4s that are triggering this fault in my collection. Or if they are, they are not a type that readily show signs of starting 5 or more seconds in.

IanSav · Post by **IanSav** » Wed Jun 15, 2016 00:59

Hi,

I have obtained a test file from Gully and tried it out on my PC, T3 and T4.

The clip is 24 minutes and 30 seconds long.

On the PC the clip always plays completely from the opening images on.
On the T3 and the T4 the clip plays from about 5 seconds into the clip.

If I try to reverse play the file the playback gets to the same point as when the video starts and then playback is stopped. I have found no way to access the first five seconds of the video.

When I perform a clean up on the clip via Fast Stream copy in VideoReDo version 5 something strange happens. When I start normal playback on the T3 and T4 the video starts playback at exactly the same point as the "uncleaned" video. However, when I reverse play the video I get to see about three seconds more of the video that I don't get to see when playing the video in the forwards direction!

I have attached MediaInfo reports for the clip before and after the VideoReDo cleanup. Apart from the version of the writing application I couldn't see a difference.

Selected T3 details are:

Drivers: 2016-04-05
Image: 4.4
Kernel: 3.6.0
Last Upgrade: 2016-06-05 01:45
GStreamer: 1.4.5

Selected T4 details are:

Drivers: 2016-04-05
Image: 4.4
Kernel: 3.14.2
Last Upgrade: 2016-06-05 01:29
GStreamer: 1.4.5

Regards,
Ian.

paullings · Post by **paullings** » Thu Jun 16, 2016 06:17

I can confirm this happens on my T3 as well.

peteru · Post by **peteru** » Thu Jun 16, 2016 09:23

Are there subtitles involved at all? Either muxed into the file or available as an external file?

IanSav · Post by **IanSav** » Thu Jun 16, 2016 09:38

Hi PeterU,

peteru wrote:Are there subtitles involved at all? Either muxed into the file or available as an external file?

There are no subtitles.

Regards,
Ian.

Grumpy_Geoff · Post by **Grumpy_Geoff** » Thu Jun 16, 2016 09:46

I observed the '+5-second' start for the first time on my T4 last night when playing a download in .mp4 format.of Happy Valley season 2 episode 1.

I then 're-encoded' (simple video/audio copy) to a .mkv container (took 2-3 seconds so I assume no conversion of the 58 min video) and the resultant T4 video playback correctly starts at the beginning (same commencement as PC).
Changed the original to a .ts container and same correct playback.

No subtitles.

[If anyone is interested, file was "Happy.Valley.S02E01.HDTV.x264-ORGANiC[ettv].mp4"]

Cheers
Geoff

IanSav · Post by **IanSav** » Thu Jun 16, 2016 10:39

Hi Geoff,

I converted Gully's clip to an MKV container (Fast Stream copy) and can confirm that the beginning of the clip is restored and available for playback.

Regards,
Ian.

peteru · Post by **peteru** » Thu Jun 16, 2016 11:17

So far all signs point to an issue somewhere in the muxing stages, based on the observation that the payload can be played back without issues if using a different container.

The most likely place to look would be dvbmediasink - I think the source has switched from sf.net to github.com between those two firmware releases. Soon after the switch there were some updates from athoik and mx3L - maybe one of those introduced the regression.

Gully · Post by **Gully** » Thu Jun 16, 2016 17:44

Just tried remuxing to the same result as the others. It plays the full file afterwards.

Another observation in case it helps, the full running time shows as the same for both the mp4 and mkv versions of the file.

Grumpy_Geoff · Post by **Grumpy_Geoff** » Fri Jun 17, 2016 16:43

Grumpy_Geoff wrote:I observed the '+5-second' start for the first time on my T4 last night when playing a download in .mp4 format.of Happy Valley season 2 episode 1.

I then 're-encoded' (simple video/audio copy) to a .mkv container (took 2-3 seconds so I assume no conversion of the 58 min video) and the resultant T4 video playback correctly starts at the beginning (same commencement as PC).
Changed the original to a .ts container and same correct playback.

Additionally, I observe a similar 3-5 second jump upon resume after pausing the MP4 file for (say) 5 seconds.
This does not occur with the 're-encoded' version in a MKV or TS container.

Cheers
Geoff

Gully · Post by **Gully** » Mon Jun 20, 2016 12:16

PeterU,

So does this provide enough info to identify and fix the cause? Is there any more testing or information needed?

peteru · Post by **peteru** » Mon Jun 20, 2016 23:01

No, not for me. I can't even reproduce the problem.

If someone else wants to dig deeper, I have provided information as to where to start looking.

Gully · Post by **Gully** » Mon Jun 20, 2016 23:16

peteru wrote:No, not for me. I can't even reproduce the problem.

Would it help if I find you a file to download that demonstrates the problem when I play it?

MrQuade · Post by **MrQuade** » Tue Jun 21, 2016 00:29

For what it is worth, I can also reproduce the reported issue on my T3 and T4 when playing back the .mp4 file specified earlier.

None of my .mp4s start out with any audio that would have been missed, so I definitely wouldn't have spotted this on any of my media.

IanSav · Post by **IanSav** » Fri Aug 19, 2016 14:58

Hi,

I had an opportunity to test Gully's file on an Enigma build using a much newer Gstreamer and the problem is no longer present. All the MP4 and MKV versions of the file played equally well.

The Beyonwiz was using Gstreamer 1.4.5 and the other box was running version 1.9.0. Both boxes were labelled as using the same kernel 3.6.0. (Gstreamer version 1.9.1 is the latest GIT version and 1.8.2 is the latest production release. The new 1.10.x production release is overdue for release so could be out any time.)

If the massive update that PeterU is reportedly working upon includes the updated Gstreamer then it is likely that this bug will be resolved.

Regards,
Ian.

Gully · Post by **Gully** » Fri Aug 19, 2016 15:17

IanSav wrote:I had an opportunity to test Gully's file on an Enigma build using a much newer Gstreamer and the problem is no longer present. All the MP4 and MKV versions of the file played equally well.

The Beyonwiz was using Gstreamer 1.4.5 and the other box was running version 1.9.0. Both boxes were labelled as using the same kernel 3.6.0. (Gstreamer version 1.9.1 is the latest GIT version and 1.8.2 is the latest production release. The new 1.10.x production release is overdue for release so could be out any time.)

If the massive update that PeterU is reportedly working upon includes the updated Gstreamer then it is likely that this bug will be resolved.

That's great to hear, Ian. Thanks for that.

PeterU, are we likely to see this version of Gstreamer in this update you are working on, and if it isn't included could you, or someone, see about updating the firmware to use this version of Gstreamer?

IanSav · Post by **IanSav** » Fri Aug 19, 2016 15:34

Hi,

I neglected to mention another difference as I didn't think it was related. I will point it out just in case it matters. The Beyonwiz uses Python 2.7.3 and the other box uses version 2.7.11.

Regards,
Ian.

peteru · Post by **peteru** » Fri Aug 19, 2016 18:18

The update I'm wrestling with is a major one. Just about everything will be updated, including GStreamer and Python.

Right now, I'm pulling in GStreamer 1.8.1, but that is likely to change before a beta is ready.

Gully · Post by **Gully** » Fri Aug 19, 2016 23:58

That sounds great, Peter.

Looking forward to it and hope the wrestling bit doesn't mean it will take too long.

IanSav · Post by **IanSav** » Mon Aug 22, 2016 15:23

Hi,

Gstreamer 1.8.3 has just been released. It is a bugfix update for the stable 1.8.x branch. It is marked as safe to overlay on any 1.8.x build.

Regards,
Ian.

IanSav · Post by **IanSav** » Mon Sep 05, 2016 05:39

Hi,

Gstreamer 1.9.2 has just been released. It is an update for the unstable 1.9.x development branch. It is expected that this will lead to a formal release of the next stable 1.10.x branch in a few weeks.

Regards,
Ian.

Gully · Post by **Gully** » Mon Sep 05, 2016 10:44

IanSav wrote:Hi,

Gstreamer 1.9.2 has just been released. It is an update for the unstable 1.9.x development branch. It is expected that this will lead to a formal release of the next stable 1.10.x branch in a few weeks.

Regards,
Ian.

Great to hear that development is progressing.

PeterU any chance of a progress report at some stage about the update you are working on? TIA

peteru · Post by **peteru** » Mon Sep 05, 2016 22:23

I wanted to release an alpha version of the next generation of firmware by now, but there are some show stoppers that need fixing before it is usable. For starters, Samba no longer works at all and so far I have not had much luck getting to the bottom of it. There are many other regressions that will need fixing.

Work continues.

Gully · Post by **Gully** » Mon Sep 05, 2016 23:09

peteru wrote:I wanted to release an alpha version of the next generation of firmware by now, but there are some show stoppers that need fixing before it is usable. For starters, Samba no longer works at all and so far I have not had much luck getting to the bottom of it. There are many other regressions that will need fixing.

Work continues.

Thanks for the update. I'm sure you will figure it out.

prl · Post by **prl** » Tue Sep 06, 2016 11:34

+1 Thanks for the update.

IanSav · Post by **IanSav** » Tue Sep 06, 2016 12:16

Hi PeterU,

Another thank you for the update. If there is anything within my capabilities with which I can help please ask.

Regards,
Ian.

About the Forum

Join us