[FFmpeg-user] framemd5 for audio?

[Paul B Mahol]

On 8/7/12, David Rice <daverice at mac.com> wrote:
>
> On Aug 7, 2012, at 7:14 PM, Paul B Mahol wrote:
>
>> On 8/7/12, Peter B. <pb at das-werkstatt.com> wrote:
>>> Hello,
>>>
>>> I'm a big fan of the framemd5 functionality for verifying losslessness
>>> :)
>>>
>>> Unfortunately, it doesn't work for checking the audio.
>>> I presume it's related to how the audio data is distributed among the
>>> frames?
>>
>> For stuff that you compare they must have same number of samples in every
>> frame.
>>>
>>> Is there any way to have framemd5 work correctly on audio, too?
>>
>> There is md5. The only way to make framemd5 output same is to make
>> both compared codecs output frames of same size - which is not possible.
>> But you could use intermediate format/codec instead....
>
> One issue here is that while there is rawvideo for framemd5 of video
> streams, there is no equivalent rawaudio format for audio data. The framemd5
> of audio will be applied to signed 16 bit raw audio (even if the source is
> unsigned or >16 bit). So for >16 bit audio the framemd5 process wouldn't
> work for losslessness verification since the bit depth would need to be
> changed prior to the application of the checksum.

AFAIK framemd5 works fine for >16 bits, just user needs to make sure that
he compares apples to apples: like same sample format (comparing s16p and s16
is possible because conversion between them is lossless), and same pcm output.