[FFmpeg-user] Why do these blank spots appear in spectrograms of converted youtube videos?

Moritz Barsnick barsnick at gmx.net
Mon Jan 9 14:26:59 EET 2017

On Mon, Jan 09, 2017 at 09:51:34 +0800, Ethan Lewis wrote:

> Thanks for suggestion but I do not have access to original audio as I
> did not upload the videos myself.

Well, you originally wrote:

> I am confused why this is happening. Is it because of ffmpeg
> conversion of m4a audio OR was it somehow fault of the original
> recording?

and it remained unclear whether you analyzed the downloaded
/path/to/download/m4a/audio or the converted test1.wav - it seemed as
if the latter, so it would make sense to analyze the former to rule out
any modification by ffmpeg. (It also remained unclear - to me - which
tool you used to create the spectrum you shared.)

It would also make sense to take another known audio speech sample of
your choice and see whether ffmpeg changes that.

By the way, when I listen to that first YouTube clip, it has obvious
heavy artefacts. I'm sure if that happens (on the "original" encoding
side), that such side effects as spectrum holes can also occur. The
heavier the compression, the more information which may be useless to
the ear is omitted - that's how the psychoacoustics of these codecs
work. (But I'm still just guessing as to where these holes *actually*
come from. I'm also pretty sure a decoder such as ffmpeg AAC/MPEG4
audio decoder would not introduce holes which were not in the source.)


More information about the ffmpeg-user mailing list