Video plays as if it was accelerated

[Bencuri]

I noticed this problem in case of my phone first, that records in 30p. But later I also noticed it when I rendered videos in Sony Vegas. And, in case of some camcorders, like the Samsung HMX S10, the raw mp4 video also had this problem.

And that is: when the video is played back, fast movements seem to be accelerated. So it seems as if the video playback would become faster for a moment. This is very strange because the timing of the video is fine anyway, it ends when it has to and the voice is in sync all through. I noticed this thing on TV as well when an amateur movie was broadcasted that won a competition. That whole movie had this problem, looked like it was accelerated all through.

In Vegas I noticed it when I rendered my videos with the Sony AVC preset. There is a MainConcept AVC preset, where I didn't notice this. Sony AVC finishes rendering much much faster than the latter one. In both cases the result is mp4, but, I don't know how, the rendering takes significantly less time with that. However those videos always have this problem I mentioned.

It is not easy to capture this problem. Maybe I am just overcomplicating this. But I wondered if you notice it. I included the original camera raw video from a footage where I had this problem, and the one rendered with Sony AVC. Let me know if you noticed this problem and if you have any idea about what causes it? Maybe an artifact of 25/30p at fast movements?

[poisondeathray]

In Vegas I noticed it when I rendered my videos with the Sony AVC preset. There is a MainConcept AVC preset, where I didn't notice this. Sony AVC finishes rendering much much faster than the latter one. In both cases the result is mp4, but, I don't know how, the rendering takes significantly less time with that. However those videos always have this problem I mentioned.

So source , and exporting with Sony AVC have problem , but MainConcept AVC does not ?? Or it does and you just didn't notice? Why don't you upload a MainConcept AVC version to compare ?

How are you viewing/ playing back these files? Maybe try something else

Maybe an artifact of 25/30p at fast movements?

You don't have 25/30p . You have 25 frames per second interlaced . If you are viewing them correctly it should look like 50p because they will be bob-deinterlaced on playback

[jagabo]

The ts file has several seconds more video than the mp4 file. They both play at the same speed for me though.

[davexnet]

I don't know the technical reason, but on my PC, frame by frame in mpc-hc on the .ts file looks good,
but on the mp4 you can see a back and forth movement suggesting a decoding error, field order problem?

On the other hand I played back both files in mpc-be and they were both fine

[Bencuri]

This problem occurs no matter if the framerate is 25 or 30, progressive or interlaced. I noticed it in all cases. I don't remember to have seen it in case of 50/60p.

Regarding playback: I target TV playback, so when you plug the USB to the TV and watch the movie like that. This is essential to me, I have to render my videos for this kind of playback, as this is the most obvious way of playback to normal users here. I have several TV-s, if the video has the problem it is visible on all. But, in case of computer playback it is also visible. I cannot mention players now, but it is irrevelant, as the TV playback from USB is the essential in this case. I couldn't eliminate the problem with trying different USB drives either, even high speed ones.

I have attached the MainConcept AVC version as well. I didn't notice the problem with that yet. I also attached a video from my phone where this problem is very visible. A band is playing and when the dancer moves fast his movements are very unnatural. To describe this better I could compare it to watching old black and white voiceless Chaplin movies. Where the video speed is faster than the natural. In case of my problematic videos it is the same feeling, but only momentarily, only on fast movements. But the weird is, the video is not speeded up at all, or should not be like that.

[poisondeathray]

I cannot mention players now, but it is irrevelant

It can be relevant, because there might be decoding issues. Also for the interlaced samples - because how it gets deinterlaced upon playback affects what you see

I also attached a video from my phone where this problem is very visible. A band is playing and when the dancer moves fast his movements are very unnatural. To describe this better I could compare it to watching old black and white voiceless Chaplin movies. Where the video speed is faster than the natural. In case of my problematic videos it is the same feeling, but only momentarily, only on fast movements. But the weird is, the video is not speeded up at all, or should not be like that.

I have a feeling you might be referring to the motion blur . This actually looks normal to me.

Faster objects will have more blur. You can influence this by shooting with a faster shutter speed (maybe not on your phone, it might not have that much control) to reduce blur, or slower shutter to increase blur.

[Bencuri]

I always use 1/50 or 1/60 as shutter speed at 25/30i or p, no matter what lighting conditions are.

By the way something came to my mind: a similar but not the same artifact could have been seen when once I converted a 30i recording to 25i. Doing it without interpolation just by using usual converter softwares, the outcome was very uneven, and the fast movements seemed accelerated, due to the missing frames I guess. After doing the conversion with interpolation in HCGUI, the outcome was proper and fast movements were much more smoother. Not exactly the same problem as the one I started the thread with, but visually a quite similar effect. That's why I have a suspicion as if it was related to a frame problem, rather than blending.

[poisondeathray]

I always use 1/50 or 1/60 as shutter speed at 25/30i or p, no matter what lighting conditions are.

But then the shutter speed and resulting blur would be inconsistent between those

For example, if you used 1/50 for 25p , you should be using 1/100 for 25i (because 25i is really 50 fields/sec) . Similarly to get the same amount of blur, if you used 1/60 for 30p , you would use 1/120 for "30i"

For general use, usually it's 1/ (2* sampling frequency)

By the way something came to my mind: a similar but not the same artifact could have been seen when once I converted a 30i recording to 25i. Doing it without interpolation just by using usual converter softwares, the outcome was very uneven, and the fast movements seemed accelerated, due to the missing frames I guess. After doing the conversion with interpolation in HCGUI, the outcome was proper and fast movements were much more smoother. Not exactly the same problem as the one I started the thread with, but visually a quite similar effect. That's why I have a suspicion as if it was related to a frame problem, rather than blending.

But if you go frame by frame, there are no duplicates or drops . So then it might be a playback problem. So then the decoding hardware/software does matter

But the band video is VFR, so that can play a role too

[jagabo]

With the 1080i videos you may be getting jerky playback if your devices aren't fast enough decode and deinterlace the video. If the decoder falls behind the display the video will stop for a frame or two, then it may skip or speed up for a few frames to catch up. This sounds like what you are describing. As an experimetnt, try encoding with no b frames. That's much easier on the decoder (though you'll need to use higher bitrates to retain quality).

Also, 30p or 30i video on a 50 Hz display will be slightly jerky under the best of circumstances.

[Bencuri]

i will test this speed problem with the decoder.

On the other hand, it came to my mind that there is an option in Sony Vegas called "Smart Resample". It is set on by default for every clip you add to the timeline. It can be changed to "Disable Resample". I always leave it on because I thought it might be useful in case of interlanced videos. But maybe I was wrong. Do you think it can also be the source of these problems in case of rendered videos?

[davexnet]

In the original file jatek.mp4 played in mpc-hc, in the lav tray icon for video settings, if you hard set bff it plays properly.
Auto or tff it looks like crap. Massive improvement in picture quality when played back at normal speed when bff is set.

The other file, 00027.ts plays back fine with auto selected as field order

[poisondeathray]

i will test this speed problem with the decoder.

On the other hand, it came to my mind that there is an option in Sony Vegas called "Smart Resample". It is set on by default for every clip you add to the timeline. It can be changed to "Disable Resample". I always leave it on because I thought it might be useful in case of interlanced videos. But maybe I was wrong. Do you think it can also be the source of these problems in case of rendered videos?

Smart resample only does something where there is a change or discrepancy in the fps (either clip, project settings, or export settings), or if you stretch/compress a clip (you're resampling the time) . It will blend frames or fields instead of duplicating or dropping

If your fps doesn't change (e.g. 25i clips, 25i timeline, 25i render, no clip stretching/shrinking) , it makes no difference



Another thing you can test playback hardware/software insufficiency problem is resizing to smaller resolution , which will consume less resources on playback. e.g. try SD PAL resolution

To rule out deinterlacing differences, you could bob deinterlace it , e.g make a 50p clip. That way you force whatever deinterlacer to be used instead of experiencing differences between HW/ SW . e.g. if one blend deinterlaces, but another uses linear interpolation deinterlacing, it's going to look very different

[jagabo]

In the original file jatek.mp4 played in mpc-hc, in the lav tray icon for video settings, if you hard set bff it plays properly.
Auto or tff it looks like crap. Massive improvement in picture quality when played back at normal speed when bff is set.

The video is definitely tff. And it plays properly as tff here.

[davexnet]

In the original file jatek.mp4 played in mpc-hc, in the lav tray icon for video settings, if you hard set bff it plays properly.
Auto or tff it looks like crap. Massive improvement in picture quality when played back at normal speed when bff is set.

The video is definitely tff. And it plays properly as tff here.

Thanks for this. I looked at the video again, this time in GraphStudio. Rendering with LAV splitter + decoder it's clearly TFF.

I seem to have something odd going on in my system. I also used the "cast to..." mechanism in Windows 10
to play it on my TV and it exhibited the same field-order stutter, while the other file was good

Sorry for muddling the thread

[Bencuri]

No problem, I welcome any notice. I also have many other problems with these files.

I will convert the jatek.ts video to DVD, and will see when played back by a DVD players what happens. Will also check this thing with the B frames.

[davexnet]

I opened jatek.mp4 in Virtualdub2. It seems to be 50 fps/ 100 fields per second - how did that happen ?

Perhaps it is not handled properly by the playback device?

[jagabo]

I opened jatek.mp4 in Virtualdub2. It seems to be 50 fps/ 100 fields per second - how did that happen ?

The source filter interpreted the field rate as the frame rate. That often happens with "separated fields" mp4 files in VirtualDub. That may be the source of the problem in the OP's players too.

[davexnet]

I opened jatek.mp4 in Virtualdub2. It seems to be 50 fps/ 100 fields per second - how did that happen ?

The source filter interpreted the field rate as the frame rate. That often happens with "separated fields" mp4 files in VirtualDub.

But stepping through it, using the viewfields filter, I see unique fields, no repetition.

Can LAV filters, ffmpegsource , Lsmash, LWLibavVideoSource, etc,etc handle properly this kind of source file?

[jagabo]

I opened jatek.mp4 in Virtualdub2. It seems to be 50 fps/ 100 fields per second - how did that happen ?

The source filter interpreted the field rate as the frame rate. That often happens with "separated fields" mp4 files in VirtualDub.

But stepping through it, using the viewfields filter, I see unique fields, no repetition.

Then it will have half the running time.

Can LAV filters, ffmpegsource , Lsmash, LWLibavVideoSource, etc,etc handle properly this kind of source file?

ffmpegSource() sees it as 50 fps and shows each frame twice, 1682 frames. LSmashVideoSource() sees it as 50 fps, 1682 frames, without repeats, but only returns data for 841 frames. LWLibavVideoSource() sees it as 25 fps, 841 frames, which is what the video really is.

[davexnet]

I opened jatek.mp4 in Virtualdub2. It seems to be 50 fps/ 100 fields per second - how did that happen ?

The source filter interpreted the field rate as the frame rate. That often happens with "separated fields" mp4 files in VirtualDub.

But stepping through it, using the viewfields filter, I see unique fields, no repetition.

Then it will have half the running time.

Can LAV filters, ffmpegsource , Lsmash, LWLibavVideoSource, etc,etc handle properly this kind of source file?

ffmpegSource() sees it as 50 fps and shows each frame twice, 1682 frames. LSmashVideoSource() sees it as 50 fps, 1682 frames, without repeats, but only returns data for 841 frames. LWLibavVideoSource() sees it as 25 fps, 841 frames, which is what the video really is.

jagabo, thanks very much your research - good info

[Bencuri]

ffmpegSource() sees it as 50 fps and shows each frame twice, 1682 frames. LSmashVideoSource() sees it as 50 fps, 1682 frames, without repeats, but only returns data for 841 frames. LWLibavVideoSource() sees it as 25 fps, 841 frames, which is what the video really is.

So it means opening it with the Libav codec in MPC-HC might improve the playback?

[poisondeathray]

So it means opening it with the Libav codec in MPC-HC might improve the playback?

Not if you were hardware limited (if the cause was slow hardware)

And it doesn't help your goal of playing it on TV / USB



Modern HDTV USB should have no problems playing any of your files. But just in case, what is the model ? You said you had several, it's unlikely that all of them would be "underpowered." What about other videos ? This only affects your batch and the single other amateur movie , correct ?

That's why I think your perception is from a shutter issue. Your 1/50 for "25i" would produce more blur than usual, and could give the perception of "moving accelerated" during fast movements . So do some tests and try 1/100 for your "25i"

In case all your TV's and your PC are both underpowered too (unlikely for *all* the TV's), an SD file should playback correctly if underpowered playback was the issue. So do some SD tests

[Bencuri]

Will check the TV specs once I arrive home in the evening.

I don't know too much about shutter speeds as I have only used 1/50 so far. Will have to test higher speeds. But one thing that makes me uncertain about it is why is it than that in case of raw, non-mp4 files I didn't notice this problem yet, in spite of that they were also recorded with 1/50?

[Bencuri]

Okay, I checked the TV types, they are both Phillips: 32PFL5604H/12 and 50PFH4009/88. The third one is lent to someone, I couldn't check that, but it is a very good Samsung. The previous two: in case of them the manual doesn't mention bitrates in the specs, but I can see they both support 50 and 60Hz playback, so fequency incompatibility cannot be the source of problems.

[poisondeathray]

But one thing that makes me uncertain about it is why is it than that in case of raw, non-mp4 files I didn't notice this problem yet, in spite of that they were also recorded with 1/50?

What about the 1st post. The original 00027.ts file ?

You said

I included the original camera raw video from a footage where I had this problem

So that implies you did notice the problem with non-mp4 files ? Or is that only after re-encoding in vegas ?

Also,

And, in case of some camcorders, like the Samsung HMX S10, the raw mp4 video also had this problem.

[Bencuri]

No, with 'I included the original camera raw video from a footage where I had this problem' I meant that this was a kind of footage where the problem occured after rendering. Sorry if it was not clear. The raw ts file had no problem. So all in all, I did not experience this in case of raw files from camcorders, except for ones that are mp4. The Samsung S10 recorded in mp4, and it had this problem. Unfortunately I don't have any of those test files now from that camcorder, so I can't remember how many FPS it was, but I guess 50i. I also have a Sanyo FH1 camcorder that records in mp4, in case of that camcorder I never experienced this problem, it is recording in 50p however. Maybe this problem relates to 25/30 p and i mp4, that's why.

Anyway I will remember your suggestion about the 1/100 shutter in the future, and will start filming like that. And will see on the long run whether this problem appears again. One thing that just frustrates me about it: I am filming with a Canon HF M41. It is very good in low light, better than some pro broadcast camcorders in that feature. But even like so, if I increase the shutter from 1/50 to 1/100, the image will be just too dark indoors. It has a gain decibel setting feature, but increasing that will add quite much noise. That's why I never went over 1/50 indoors. We are doing a lot of videos in clubs, 1/100 shutter will cause a lot of trouble for me. And I cannot consider buying another camera in this price range, because as I said, even pro ones in this range are worse in low light. I could buy a 3-4000$ one and it would record noisy image indoors. At least this camcorder that I have records fine in low light. Going over that price is not an option for me at the moment.

[Bencuri]

With the 1080i videos you may be getting jerky playback if your devices aren't fast enough decode and deinterlace the video. If the decoder falls behind the display the video will stop for a frame or two, then it may skip or speed up for a few frames to catch up. This sounds like what you are describing. As an experimetnt, try encoding with no b frames. That's much easier on the decoder (though you'll need to use higher bitrates to retain quality).

Also, 30p or 30i video on a 50 Hz display will be slightly jerky under the best of circumstances.

Unfortunately Vegas doesn't offer rendering mp4 without B frames. I dried to export it as mpg, that allows it, but for some reason I get an error message if I start rendering like that. The only way I could to it is rendering it to HDV transport file, but that type is not supported by my TV.

[jagabo]

The only way I could to it is rendering it to HDV transport file, but that type is not supported by my TV.

HDV is MPEG 2 video. Try remuxing to an MPG container. Your TV might support that.

[davexnet]

I got a DVD-spec MPG using AVStoDVD/HCenc to encode jatek.mp4 that looks OK
I used this script to feed the file to the program:

loadplugin("d:\Program Files\AviSynth\plugins\LSMASHSource.dll")
V = "C:\Users\davex\Downloads\jatek.mp4"
LWLibavVideoSource(V)

and the program generates this script to pass to the encoder (after adding the audio and manually flagging the source as interlaced) :

LoadCPlugin("I:\program files\AVStoDVD_287_NoInstall\Lib\ffms2.dll")
LoadPlugin("I:\program files\AVStoDVD_287_NoInstall\Lib\LeakKernelDeint.d ll")
Audio = FFAudioSource("C:\Users\davex\Downloads\jatek.mp4" , track=-1)
Video = Import("C:\Users\davex\Desktop\scr1.avs")
Video = Video.ConvertToYV12(interlaced=true)
Video = Video.LeakKernelBob(1,7,false,false)
Video = Video.blackmanresize(720,576)
Video = Video.SeparateFields().SelectEvery(4,1,2).Weave() # Or Video=Video.Assumetff().SeparateFields().SelectEve ry(4,0,3).Weave()
AudioDub(Video, Audio)


EDIT added a script comment

[poisondeathray]

No, with 'I included the original camera raw video from a footage where I had this problem' I meant that this was a kind of footage where the problem occured after rendering. Sorry if it was not clear. The raw ts file had no problem.

Sorry , my misunderstanding


It might be your TV is not expecting interlaced files in MP4 container. Or that the field rate metadata when you use mp4 might be confusing your TV. Or just a bug with the way Sony exports interlaced AVC into MP4

Frame rate mode                          : Constant
Frame rate                               : 50.000 FPS
Original frame rate                      : 25.000 FPS

Then you can render it as the AVCHD (50i) preset. This will put it into a transport stream (.m2ts) with the correct 25 fps with no ambiguity. The one "negative" of transport streams is they have more overhead, typically ~6-7%

Either that, if you need MP4 container, bob deinterlace it to 720p50 , which most TV's that accept AVC L4.1 and USB input will be able to playback