Re: [Mlt-devel] preserving source video quality (-sameq?)

[Dan D. <da...@de...>]

On Mon, Feb 14, 2011 at 10:10 AM, Bogdan Licar <ree...@gm...> wrote:
> I have to use melt, since this is part of a web app dealing with video
> editing.

But you could compare your experience with using a frontend that
already embodies quite a bit of knowledge to see how it compares with
your experience trying to learn melt.

> Anyway, I'm dealing with variable frame rate videos, so maybe that's the
> problem.
>
> Here you can download a sample clip I'm working on (500KB)
> http://www.mediafire.com/?px20612m82idsu4
> When I play it with "melt clip.flv" the video is too fast and the audio out
> of sync.
> In order to play it at the correct speed I have to use a profile, otherwise
> I don't know any other way to do that.

More often than not you _should_ specify a profile especially if you
are executing a composition. More than likely, without a profile and
with a variable fps, melt detects and tries to use some strange fps.

> It plays at the correct speed with frame_rate_num=25 frame_rate_den=4, but
> this obviously has a crappy fps.
>
> If I join it with a video with the same codec and stream settings, I get
> fast forwarded videos and out of sync audios,
> with most of black screen, since, for instance, for the clip above -consumer
> xml returns out=133, but the output clip ends way before that.
>
> Really stuck here, perhaps converting it all to a constant frame rate could
> solve the problem.

I already told you about the state of support for variable fps.

> On Mon, Feb 14, 2011 at 6:48 PM, Dan Dennedy <da...@de...> wrote:
>>
>> On Mon, Feb 14, 2011 at 8:50 AM, Bogdan Licar <ree...@gm...>
>> wrote:
>> > Hi Dan,
>> > which is the best way of dealing with videos with different framerates,
>> > when
>> > joining them together?
>>
>> It is automatic. All inputs are conformed to the output framerate
>> expressed by the profile or a consumer property overriding the profile
>> fps.
>>
>> > I'm not talking about the troublesome videos, but clean ones. Should I
>> > convert them all to one fps, and then join them?
>>
>> no
>>
>> > Because I'm having problems here, I don't know which
>> > profile settings to use when
>> > reading the framerates from the xml consumer and when joining them
>> > together.
>>
>> I do not understand, and you have not provided much information. You
>> use the profile settings that you want to use, which is not
>> necessarily based on the inputs. When authoring MLT XML or using melt
>> command line syntax you express timing in frame units based on the
>> output fps. Maybe you should try using a gui frontend like OpenShot or
>> Kdenlive.
>>
>> > On Tue, Feb 8, 2011 at 10:26 PM, Dan Dennedy <da...@de...> wrote:
>> >>
>> >> On Tue, Feb 8, 2011 at 1:06 PM, Bogdan Licar <ree...@gm...>
>> >> wrote:
>> >> > Thanks for the reply :)
>> >> >
>> >> > I'm using melt for joining plenty of clips with fade in, fade out,
>> >> > and
>> >> > watermark effects on them.
>> >> > So I pretty much need it, there's nothing else that good.
>> >>
>> >> OK, I was under the wrong impression.
>> >>
>> >> > What I do now is -consumer avformat:video.mp4 qscale=.1, and I manage
>> >> > to
>> >> > preserve
>> >> > source video quality, but I still have the speed problem - the sound
>> >> > plays
>> >> > at the correct speed though.
>> >> >
>> >> > If I set frame_rate_num=1 and frame_rate_den=1, the speed is correct,
>> >> > but I
>> >> > obviously get only 1
>> >> > frame per second.
>> >> >
>> >> > So I've checked and mediainfo says it's a variable framerate, as you
>> >> > guessed.
>> >> > Then I've converted my clip to a constant 25 framerate, with ffmpeg's
>> >> > -r
>> >> > option.
>> >> > Here's the new xml output (the stream frame rate is 25, but the
>> >> > codec's
>> >> > frame rate is still 1000,
>> >> > dunno what this actually means).  http://pastebin.com/UxpQjbVM
>> >>
>> >> Did you use -vcodec copy? If so, then I do not think it actually drops
>> >> and repeats frames as needed. It just set the data in the container
>> >> file to indicate 25 fps.
>> >>
>> >> > I'm working with videos recorded from various webcams and with this
>> >> > particular video - with the constant
>> >> > framerate - I still can't get the right speed (although I get very
>> >> > close
>> >> > to
>> >> > it when I set frame_rate_num=5, frame_rate_den=1;
>> >>
>> >> I have had some good results converting troublesome inputs to a
>> >> lossless matroska:
>> >> ffmpeg -i the.flv -vcodec huffyuv -acodec flac out.mkv
>> >>
>> >> If you want to output from melt in that format: melt ... -consumer
>> >> avformat:out.mkv vcodec=huffyuv acodec=flac
>> >>
>> >> > with a lower numerator, it won't play at all). But with other webcam
>> >> > recorded videos changing to constant framerate and
>> >> > setting a profile with a 25 framerate ratio works. So this particular
>> >> > clip
>> >> > is somehow corrupted - timestamps issues I guess.
>> >> >
>> >> > thanks again Dan
>> >> >
>> >> > On Tue, Feb 8, 2011 at 7:41 PM, Dan Dennedy <da...@de...> wrote:
>> >> >>
>> >> >> On Tue, Feb 8, 2011 at 5:02 AM, Bogdan Licar
>> >> >> <ree...@gm...>
>> >> >> wrote:
>> >> >> > I've tried converting my videos using libx264 - recorded with the
>> >> >> > webcam
>> >> >> > -
>> >> >> > but they have
>> >> >> > some problems (bad timestamps), but I've managed to convert them
>> >> >> > to
>> >> >> > other hq
>> >> >> > flash videos.
>> >> >> >
>> >> >> > But now I have the profile problem: I can't seem to get the
>> >> >> > correct
>> >> >> > framerate in order to
>> >> >> > play the video with melt at the correct speed (it's always too
>> >> >> > fast).
>> >> >> > Here's
>> >> >> > the -consumer xml
>> >> >> > output on a 28second video http://pastebin.com/JV50kAbd
>> >> >> > I've tried frame_rate_num=25 and frame_rate_den=1 and other
>> >> >> > combinations,
>> >> >> > but it's always too fast.
>> >> >>
>> >> >> I can understand how you may think it is a profile problem but it is
>> >> >> not. Theoretically, you can change the profile framerate at will and
>> >> >> the framework will conform all inputs to it. Therefore, a speed
>> >> >> problem is almost always due to the producer misbehaving -
>> >> >> ffmpeg-based avformat in this case. In looking at your XML, I see it
>> >> >> is reporting a fps of 1000. I usually see that with variable
>> >> >> framerate
>> >> >> clips, and unfortunately, MLT does not handle variable framerate
>> >> >> very
>> >> >> well at all. So, at this time I tell people that is not supported,
>> >> >> and
>> >> >> I am personally not going to look into improving that any time soon.
>> >> >> A
>> >> >> simple way to confirm the variable framerate is with the mediainfo
>> >> >> command.
>> >> >>
>> >> >> Sorry to be such a bummer on this thread. I would say based upon
>> >> >> your
>> >> >> goals to simply cut flv, no image filtering, a desire to use sameq,
>> >> >> and variable inputs that MLT is a very bad fit for you. You may want
>> >> >> to stick with the ffmpeg command line or maybe avidemux will work.
>> >> >>
>> >> >> > Thanks again, Dan
>> >> >> >
>> >> >> > On Mon, Feb 7, 2011 at 9:14 PM, Dan Dennedy <da...@de...>
>> >> >> > wrote:
>> >> >> >>
>> >> >> >> On Mon, Feb 7, 2011 at 11:42 AM, Bogdan Licar
>> >> >> >> <ree...@gm...>
>> >> >> >> wrote:
>> >> >> >> > Hi Dan,
>> >> >> >> > so how could I export to a high quality flv? With the help of a
>> >> >> >> > profile?
>> >> >> >> > Because right I'm getting a low quality output flv, and I've
>> >> >> >> > also
>> >> >> >> > tried
>> >> >> >> > using
>> >> >> >> > a profile with the same settings as the source video.
>> >> >> >>
>> >> >> >> MLT profile has nothing to do with encoding quality. The only way
>> >> >> >> to
>> >> >> >> get good quality FLV is to use the H.264 video codec because the
>> >> >> >> Flash
>> >> >> >> 6 (Sorenson) video codec sucks no matter what you try to do. Of
>> >> >> >> course, it is all relative. Also, of course, if you using Flash 6
>> >> >> >> FLV
>> >> >> >> as input, then garbage in is garbage out. Do you know the ffmpeg
>> >> >> >> parameters not using -sameq to encode your idea of "high qulity
>> >> >> >> flv?"
>> >> >> >> If so, then use them in melt after the "-consumer avformat". If
>> >> >> >> you
>> >> >> >> do
>> >> >> >> not know then search the web for suggestions or look at what
>> >> >> >> tools
>> >> >> >> like Kdenlive, WinFF, and other FFmpeg frontends are using - the
>> >> >> >> ones
>> >> >> >> with which you have had good results.
>> >> >> >>
>> >> >> >> > Thanks
>> >> >> >> >
>> >> >> >> > On Mon, Feb 7, 2011 at 7:32 PM, Dan Dennedy <da...@de...>
>> >> >> >> > wrote:
>> >> >> >> >>
>> >> >> >> >> On Sun, Feb 6, 2011 at 8:31 PM, Bogdan Licar
>> >> >> >> >> <ree...@gm...>
>> >> >> >> >> wrote:
>> >> >> >> >> > Hi,
>> >> >> >> >> > I'm joining some flv videos into a single one, but the
>> >> >> >> >> > output
>> >> >> >> >> > video
>> >> >> >> >> > has
>> >> >> >> >> > a
>> >> >> >> >> > much lower quality than the various clips.
>> >> >> >> >> > In ffmpeg I can add a -sameq parameter, and the quality is
>> >> >> >> >> > preserved,
>> >> >> >> >> > but
>> >> >> >> >> > this doesn't work in melt when I add
>> >> >> >> >> > it at the end of the melt command.
>> >> >> >> >> >
>> >> >> >> >> > Is there a way to make melt preserve source video quality?
>> >> >> >> >>
>> >> >> >> >> melt does not support -sameq.
>> >> >> >> >> In MLT, everything is brought to uncompressed and re-encoded,
>> >> >> >> >> if
>> >> >> >> >> necessary.
>> >> >> >> >>
>> >> >> >> >> --
>> >> >> >> >> +-DRD-+
>> >> >> >> >
>> >> >> >> >
>> >> >> >
>> >> >> >
>> >> >>
>> >> >>
>> >> >>
>> >> >> --
>> >> >> +-DRD-+
>> >> >
>> >> >
>> >>
>> >>
>> >>
>> >> --
>> >> +-DRD-+
>> >
>> >
>>
>>
>>
>> --
>> +-DRD-+
>
>
> ------------------------------------------------------------------------------
> The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE:
> Pinpoint memory and threading errors before they happen.
> Find and fix more than 250 security defects in the development cycle.
> Locate bottlenecks in serial and parallel code that limit performance.
> http://p.sf.net/sfu/intel-dev2devfeb
> _______________________________________________
> Mlt-devel mailing list
> Mlt...@li...
> https://lists.sourceforge.net/lists/listinfo/mlt-devel
>
>



-- 
+-DRD-+