[Ffmpeg-devel] Re: [PATCH] PKT_FLAG_B

[Baptiste Coudurier]

Hi

Michael Niedermayer wrote:
> Hi
> 
> On Thu, Jul 20, 2006 at 12:24:01AM +0200, Baptiste Coudurier wrote:
>> Michael Niedermayer wrote:
>>> Hi
>>>
>>> On Wed, Jul 19, 2006 at 11:52:38AM +0200, Baptiste Coudurier wrote:
>>>> Hi
>>>>
>>>> Here is a patch that define PKT_FLAG_B to mention that a packet contains
>>>> a B frame.
>>>>
>>>> That's not much yet, but it will work with stream copy and it's very
>>>> useful. Is that acceptable ? If you think about another name, Im fine also.
>>> for what is it usefull?
>>>
>>> [...]
>> Mainly because the pts and dts difference is sometimes not exact. I got
>> samples here, in which P frames have pts/dts difference of 1 like a
>> delay or something.
>>
>> Actually there are at least 2 muxers that would need that (mov, and
>> gxf), dunno any other muxer need that..
> 
> could you elaborate on what these muxers do with the frame types / why
> they need them?
> 
> [...]
> 

GXF muxer mark the frame as being a B frame. That fix the seeking on
broadcast server (direct seek, no blocky), and permits indexing on import.

MOV muxer, it's not mandatory, but It will help deciding whether to put
'ctts' atom or not. Like I said I got some mpeg samples with P frames
only and those don't need ctts atom, however detection based on pts/dts
is not really accurate since difference is still 1 and that will create
'ctts' atom.

Also I would say that is mainly useful during stream copy, pts/dts
during encoding are accurate.

-- 
Baptiste COUDURIER                              GnuPG Key Id: 0x5C1ABAAA
SMARTJOG S.A.                                    http://www.smartjog.com
Key fingerprint                 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA
Phone: +33 1 49966312