Proposal Beat Finder
|Proposal pages help us get from feature requests into actual plans. This page is a proposal for graphically showing the beats in a track.|
Proposal pages are used on an ongoing basis by the Audacity development team and are open to edits from visitors to the wiki. They are a good way to get community feedback on a proposal.
- Note: Proposals for Google Summer of Code projects are significantly different in structure, are submitted via Google's web app and may or may not have a corresponding proposal page.
Find periodic beats in a piece of music. Show these graphically in a track. Allow users to more easily time-shift audio so that the beats line up.
- Peter 26Jan19: Note that this can be achieved with the Queen Mary Vamp plug-ins
- None so far. The Nyquist list is the best place to try as there is a nyquist beat finder plug-in.
- Makes for good crossfades when transitioning between songs.
Could it be possible to improve the beat finder? It usually works quite good on some parts of the sound (usually the verse and the ref), but breaks, bridges, and other rhythmically complex stuff obfuscates it. Well ... sometimes that obfuscates that even human beings, and that's why musicians keep taping their feet to stay on the beat. Also the beat finder sometimes find different measures in refrains and verses due to different dynamics and drum patterns.
That brought me to the following idea: Could the Beat Finder get extended to more musical tool like a "Song Structure Assistant"?
- Analyse the found beats for any strong periodicity. This will probably have something to do with the beats, present this to the user with an example where these beats have a strong signature in the song.
- Let the user make adjustments like the detected beats are just halftime (e.g. the beat finder has just found the snare drum on 2 and 4 of the bar) and add information like the time measure (4/4, 3/4, ...)
- Use this information to redo the beat finding and this time filter out "double beats" by probability how good they fit into the time signature and interpolate during parts where no clear beat could be found in the first run. Use that information to find variations in the timing, like songs recorded without a click track or life recordings.
- Use this information and the knowledge that verse, refrains, ... are usually multiples of 4 beats long and have different dynamics to identify the musical structure of the song.
- Ok, that's now a little bit sophisticated and maybe un-necessary: Verify and refine this by searching for correlations of the envelopes of the structures found is step four.
- Let the user confirm or edit and name these structures.
If that works one could work in a much more musical way with Audacity like: Setting the mark on "count time 3+ of the fourth bar in second verse". Another fine use would be to export this list of beats in an format usable by video editors to automate that kind of video clip "on the beat cutting". Or "straighten" a life recording with an unsteady timing by feeding the differences to an accurate timing to the time stretcher. Or nearly automate remixes and rearrangements, or ... I see a lot of useful applications once the musical structure of a song is known.
Maybe this request could be split up in two: "Improve the beat finder" (Step 1-3) and "Analyse for musical structures" (Step 4-6)