I have read several posts about the drifting tempo in MP3s and I am having similar issues trying to set a drum sample track along with the original recording. Unsurprisingly, the original musicians drift somewhat about the average tempo. It seems like the Time Track function is what I need, but adjusting the time curve over the course of the song seems a bit awkward. I think an easy solution would be to tap out a click track (using the keyboard or mouse?) along with the original recording following along with that beat, automatically taking an average tempo from the click track, creating a Time Track from the average and remaining residual beat variations in the click track, then using the Time Track to do a continuous tempo adjustment. By chance, has this already been done and my forum searches have simply not been thorough enough?
That’s one of those tools that looks good on paper. In order to get rhythm management like that to work, you need to know what the show rhythm is now and that’s not easy. If you have a song where the drum or rhythm guitar come and go through the course of the show, then the program has to average long portions of the show to get the BPM. That’s precisely what you can’t do if the performed rhythm moves about.
So the first time the song takes a phrase “break” like many dance tunes do, that’s the end of the tool. If there’s enough of those, you’d be doing much of the cleanup manually anyway.
I re-read my post and realized that I may not quite have captured exactly what I was thinking. I was thinking that the user could create the click-track by following along as the song plays all the way through (maybe I used the phrase “click-track” in the wrong context here). Automation would take over from that point with the user-defined click-track as a tempo reference, rather than trying to recognize the beat itself. So the results would be entirely dependent on the user’s ability to keep the performers’ beat.
You’re right, though, substantial rests would have to be masked out of the averaging and tempo correction, and even then, a good deal of editing may still be required. Even humans that keep a good beat would have a tough time tapping out more than a few beats worth of silence, especially when the performers were doing the same thing.
Thanks for your reply. It gives me more things to consider. “I can dream up schemes when I’m sitting in my seat!”
What you are describing is a type of “Tempo Mapping” by which logical time (bars and beats) is mapped against real time (hh:mm:ss).
This is related to (but different from) a suggestion on the Feature Requests page of the wiki for “snapping to beats”. (search that page for Snap to).
When you’ve had time to “dream up a scheme” you may want to post your idea(s) on the “Adding Features” section on this forum for discussion and possible promotion into future versions of Audacity. http://forum.audacityteam.org/viewforum.php?f=20