[mythtv] [PATCH] time stretch with libsoundtouch

Jesper Sörensen jesper at datapartner.se
Tue Nov 2 11:17:14 UTC 2004


>there is no problem for libsoundtouch and libsamplerate coexisting. this is
>how I have implemented it.
>libsoundtouch does not even get instantiated until the first non 1.0
>stretchfactor set. thus if you dont use it it has no impact. this should
>satisfy you I hope.
>  
>

Sounds good to me.

>>I haven't tried Mark's patch yet but I'm planning to. Anyone know how
>>much one can "bend" the audio without the chipmunk effect getting too
>>bad? I'm thinking that a couple of percent change in speed wouldn't
>>affect the pitch that much. I've been planning to try and reimplement
>>the audio warp for the AV sync using either libsoundtouch or
>>libsamplerate, but I don't know when I'll find the time. I want to take
>>    
>>
>there is no lipsync problem. I have catered for that by estimating the
>amount of sound buffered and calculating the audio timecode as a result so
>it matches the warped audio to the video timecodes. seems pretty good.
>there is no chipmunk effect. +25% is usable even +50% if they talk a bit
>slow. as I said before, I can watch stargate eps at +25% quite easily and it
>seems almost normal. they just talk fast.
>  
>

Do you happen to know how much the audio can be warped with 
libsamplerate (without pitch correction) before people start to sound 
like chipmunks? (Or doing "whale" talk like in Finding Nemo.) I was 
wondering if I could get by without libsoundtouch (need to save some CPU 
cycles). I guess I'll have to do some experimenting...

BTW, my problem is not that I need to speed Stargate up, I need to slow 
them down to make it last longer. I'm jonesing pretty bad for a new 
episode but if I recall correctly there's no new SG-1 or Atlantis until 
January or so. :'(



More information about the mythtv-dev mailing list