Re: [Mlt-devel] flv - small sync, minutes of invalid

[Carl K. <ca...@pe...>]

On Sat, Nov 21, 2009 at 12:23 AM, Dan Dennedy <da...@de...> wrote:
> On Tue, Sep 29, 2009 at 10:36 AM, Dan Dennedy <da...@de...> wrote:
>> On Mon, Sep 28, 2009 at 9:56 PM, Carl Karsten <ca...@pe...> wrote:
>>> On Mon, Sep 28, 2009 at 2:01 PM, Dan Dennedy <da...@de...> wrote:
>>>> On Sun, Sep 27, 2009 at 2:45 PM, Carl Karsten <ca...@pe...> wrote:
>>>>> I made a red/blue test flv, uploaded to blip, it plays perfectly:
>>>>> http://carlfk.blip.tv/file/2650272/    nothing new here, but if you
>>>>> want the details, I am setting the description to the melt and .mlt.
>>>>> I am not escaping the xml right, so it doesn't show up, but it's in
>>>>> the source.
>>>>>
>>>>> So I used the 'same thing' on a talk video.  there is a slight sync
>>>>> problem about 3 seconds after 20 min, and
>>>>
>>>> I wonder if you are running into a problem with DV unlocked audio. You
>>>> can search to learn more, but Kino users had similar problems with
>>>> footage captured from certain devices. Then, I added some special
>>>> handling in Kino for this problem, and it resolved nearly all
>>>> complaints. So, to test this hypothesis, you can load the source DV
>>>> file into Kino, do no editing, and export it as a new DV file leaving
>>>> the resample audio checkbox enabled. This process converts it from
>>>> unlocked to locked and uses resampling to interpolate the missing PCM
>>>> data. Then, use the output of that in your mlt and script and see if
>>>> resolved it.
>>>
>>> Well, you are onto something.   It did get rid of the audio sync
>>> problem when blip converted it.
>>
>> I am adding near the top of my MLT ToDo list, to work on better
>> support for unlocked DV audio. I can not jump on it right away, but it
>> is rather important. I am not sure I can make it automatic. I might
>> have to make it something like Replay Gain normalization. That means
>> you would do two passes: the first pass will report a correction value
>> that you supply to a filter for the second pass.
>>
>
> Carl, I just found and fixed a math bug that was causing the MLT
> avformat producer to drop some extra audio samples. This would account
> for some a/v sync drift as well as some crackling in the output. There
> will be a new release by the end of the month.

Thanks for keeping me posted.


-- 
Carl K