how to mash up
this is for full-track beatmatching, or changing the length of a whole song to match another. there are absolutely other techniques to make mashups, and automated processes, but this is the old fashioned way, kind of like stretching out two tape reels until they’re exactly the same length. (or, until the measures are the same length, or until the “beats match,” if for instance the songs have a different verse/chorus structure.)
what you need:
files of each song
key and bpm of each song (ideally)
audacity (it’s free)
strength of will to calculate a percentage and/or be wrong
let’s say you have two songs, one that is 139bpm and GMajor, and one that is 170bpm and Eminor. let’s gloss over the fact that Em is a sympathetic minor to GM, so we don’t have to change the pitch, and let’s not worry too much about music theory, or how to immediately know that, because in most cases you can just change the pitch of one track by +1 or -1 until you find a good match.
drop the tracks into audacity. you can do this in any order you want, but make sure you apply transformations to the correct track. the most important thing you need to know about audacity is these little fuckers!!
the cursor icon selects portions of a track. the two arrows move tracks around. i don’t know what the others do and honestly fuck them, but if you don’t make sure have the right type of “select” selected, you’re going to struggle. for most of this tutorial you will need the double arrow, because we’re moving tracks around to match the beats, but if you want to select a portion to cut and paste (see end) you will need the cursor.
for most of this tutorial, you want to have only one track selected, but the entire track. to do this, click the empty area below the volume/pan sliders on the left.
you need to either speed up the slower track, or slow down the faster track. i think you can use the same number for both but i might be wrong about that because i usually have to do a lot of fine tuning. this could also be because key/bpm sites are not always accurate. you can always try tapping out the tempo with a tap metronome, but i usually just proceed with guesswork.
let’s slow down the faster track. i just google “139 is what percent of 170″ (smaller tempo is what percent of bigger tempo), and i use geteasysolution because it provides enough decimal points (we need 3). the answer we’re looking for is 81.764. subtract that from 100. 100 - 81.764 = 18.236. that’s all the math, i promise.
select the faster track (if you don’t select a track, audacity will transform all tracks, so be careful), and select effects > change tempo. (don’t select “change speed” because that will change the pitch as well, which you could do, but it’s much harder.) enter the number we calculated above, and because we’re slowing down the faster track, make it a negative number:
if you wanted to speed up the slower track instead, use the same number, but as a positive value. (i could be wrong about this, and maybe you should calculate the percentage differently depending on which transformation you want to make, but really any number that’s close enough is good enough.) you can can change either track, or a little bit of both (cut the number in half, speed up the slower track by the positive half-number, and slow down the faster track by the negative half-number). however, some songs are occasionally just too gosh darn different to mash up without degrading the audio quality. sometimes you have to double or halve one of the tempos before you start calculating the percentage. (i.e. for 80 & 180, you might want to calculate “80 is what percent of 90,″ depending on how you want the measures to match.)
at this point you probably need to zoom in at the very beginning and match the first downbeat, unless you’re very lucky and they happen to occur at the same point in each file. the first downbeat might not be the first spike in the waveform, so you might need to mute each track, and look at and listen to each individually, until you can tell which spots need to vertically align. i’ve added red lines (in mspaint, audacity doesn’t have red lines) surrounding the alignments we’re looking for:
they are not always obvious and it’s easy to lose hours doing this. these are the first beats of each measure and (because both songs are in 4/4 time signature, which most but not all songs are), you could count them to ensure there are 4 beats per measure in each track and they’re matching up correctly, but i usually just sort of go by feel more than by precisely counting beats.
if they’re a little bit out of alignment, you need to figure out which one’s faster and slower, and make small adjustments. it’s easy to hear and see, over the course of a four minute song, a difference of as little as 0.01%, but it’s not always easy to make the correct adjustments. sometimes i change a song, in tiny increments, by as much as 2% or something, in addition to the original number calculated. i know i said there wasn’t any more math because this is, uh, applied, audio engineering?
there are any number of ways you can match the beats, and none are necessarily more correct than any other, but you might want to zoom out and look for the biggest spikes and drops in the waveform and try to match those. these usually correspond to verse/chorus transitions, and not all songs have verses and choruses that are the same length, but more often than not, if you get the tempo right, it sort of falls into place, even if the larger structure is made of different shapes.
let’s assume for the purpose of this tutorial you’ve picked two songs that have the same verse/chorus structure, that are the same time signature, that don’t have any tempo or key changes -- although you can work around all of these things using the same basic tools (cut and paste, change pitch, change tempo)
i’m really sorry if these verse/chorus transitions aren’t immediately obvious, but it gets easier after you make like 150 of them, i promise. you can go back and listen to this mashup to compare if you like, and you can even probably form a reasonable argument that i’ve put them in the wrong place, and the mashup would be better if i rearranged them. you might be right. you might also notice there’s 30 seconds of silence at the end of the bottom track. the actual mashup has a third track (tracks > add new > stereo track), where i copied & pasted the first 30 seconds of the shorter track to fill the gap:
because the 21 pilots track just sort of cuts off at the end, i was able to pretty easily just cut it off at the end of the measure, but for other songs you might need to add a fade out (also available in the effects menu).
you probably want to get the tempo transformations finished before you start doing any cutting or pasting, because if you have tracks split up into multiple pieces, or if you try to apply transformations to multiple tracks, they may not maintain the alignment you had before you applied a multiple transformation. audacity works best if you keep the track whole; if you want to split it up into 64 pieces for each 2 bars and apply transformations after dropping markers, use a different program. audacity’s split track (cursor-type select + ctrl + i) does not work like markers. ableton would be great for that kind of control, and you can get a free 30 day trial, or if you’re willing to fork over $99 you can get the basic version (i did, but i don’t use ableton for mashups).
in order to match pitch (and really most of the time you should do this first, but the example in this tutorial was an example that already matched pitch), it can get complicated quickly but basically the rule is, if both songs are in major, or both or in minor, the letter needs to match. (i.e. for GMajor and CMajor, you need to change them to be the same letter. you can raise G +5 semitones or half-steps (the unit audacity uses in effects > change pitch) to match C, or you can lower it -7 semitones. you can raise C +7 semitones to match G, or you can lower C -5 semitones to match G. this is actually a bad example because that might be too large a transformation to avoid audio distortion, and if the two songs were that far apart i might split the difference.
if one song is in minor and one song is in major, you need to make the major one’s letter be two later than the minor one. in the example here, Eminor is two earlier (in the alphabet) than Gmajor. Aminor is two earlier than Cmajor. Emajor is two later than... surprise. C sharp (#) minor. good luck! learn the circle of fifths lol. (again, close enough is good enough to start fine tuning -- if your pitch is 1 semitone off it will sound obviously wrong, and you can try some +1 or -1 guesses until you find something that sounds right.)
audacity exports to .wav -- i use an online mp3 converter but you can also download an... encoder library...? to make audacity capable of exporting to mp3
audacity also might be how some of the (bass boosted) or (playing from another room) remixes are made. i also have an entire blog dedicated to simply slowing songs down, which never caught on, but i did all of them in audacity too
