j'ai trouvé l'explication du probleme dans l'aide de virtual dub --->
par contre ça ne dit pas si la duréee est est celle affichée ou non. ( pour le mp3 a taux variable , c'est clair que non !! car 6min pour un morceau de 3 ça ce sentirais je penses ....)
Fraunhofer-IIS's MPEG Audio Layer 3 codec
There are three versions of Fraunhofer-IIS's MP3 codec:
· Standard version.
This version will decode all MP3 streams, but cannot compress. Newer versions of Windows Media Player come with this codec.
· Advanced version.
The Advanced codec will decompress all MP3 streams and will also encode MP3 streams up to 64 kilobits/sec (22KHz). It usually comes with Microsoft Netshow.
· Professional version.
Decodes all MP3 streams and encodes up to 128 kilobits/sec (CD quality). This codec is very hard to find, and its distribution outside of a product is… well, "questionable."
Most people do not have the Professional version, and this is the reason why you can often play an AVI file with a 128Kbps audio stream, but can't make one yourself. (It is technically possible to encode an MP3 with another program and directly incorporate it into an AVI with the correct tag, but I haven't been able to do it yet.) Note that since the audio stream is a valid MPEG Layer 3 stream, you can extract the audio track in its entirety with VirtualDub and play it with an MP3 player such as WinAmp.
Another problem with this codec is that it is an enormous hack. In particular, it sets the nBlockAlign member of the wave format to 1. This means that applications think a single byte can be decompressed into audio, when in fact the Fraunhofer-IIS codec will buffer up data until it has enough to decompress a Layer 3 frame. Extracted sections of such audio streams will often have muted tails because the audio codec discards fractions of frames at the start. The root of these problems is the MP3 spec, which allows audio frames to "borrow" unused bandwidth from earlier frames, making it impossible for the MP3 audio codec to specify a fixed size for a self-contained audio block.
Finally, the Fraunhofer-IIS codec is very lazy and incorrectly sets the bitrate of the stream. For instance, a 48Kbit stream encoded by the Fraunhofer-IIS codec has a specified rate of 6000 bytes/sec, when in reality the stream is about 5971 bytes/sec. This 0.0048% difference may not seem like much, except that it causes the audio to race past the video approximately one second for every 200 seconds of video. One way to ‘fix’ this problem is to correspondingly adjust the video frame rate to compensate. VirtualDub will automatically correct for this problem when compressing audio to MPEG format.
