One question, three different jobs
When people ask how to change the pitch of a song, they usually mean one of three different jobs. Figuring out which job you actually have is most of the battle, because each one uses a different process, and each process costs you something different in sound quality.
Job one is transposing: shift the key so you can sing along comfortably, without touching the tempo. Job two is varispeed: pitch and speed move together, the way a tape machine or a turntable does it. Job three is slowing a song down for practice without the pitch dropping along with it.
We built a free Pitch Changer that handles all three. It runs in the browser, nothing uploaded, we never see the file. The rest of this post is about picking the right job and the right settings.
Semitones, in plain terms
A semitone is the smallest step in Western music: the distance from one piano key to the very next one, black keys included, or one fret on a guitar. Twelve semitones make an octave.
When a tool asks how many semitones to shift, it is asking how many of those steps to slide every note in the song. Shift up 2 semitones and a song in C major becomes a song in D major. Shift down 3 and G major becomes E major. The melody, the chords, and the relationships between notes all stay intact; the whole thing just sits higher or lower.
As a rough guide: 1 or 2 semitones is subtle, 3 or 4 changes the feel of the song noticeably, and past 5 in either direction most material starts to sound processed. That is not a rule, just where the artifacts tend to become obvious.
Job one: transpose it into your singing range
This is the karaoke and cover-song job. The original singer's chorus sits just out of reach, so you drop the whole song 2 or 3 semitones and suddenly the high note is yours. The speed stays exactly the same, so you can still play or sing along with the original arrangement.
Under the hood, the tool has to move every frequency in the mix while keeping the timing untouched. That math is good these days, but it is not free. Small shifts of 1 to 3 semitones sound close to transparent on most material. Bigger shifts start to warble, especially on dense mixes and exposed vocals. One useful trick if you need a big move: raising a song 4 semitones puts it in the same key as dropping it 8, just an octave apart, and the smaller of the two shifts almost always sounds cleaner.
It helps to know your starting key before you count semitones. Drop the track on our free Key & BPM Finder, or read our guide on how to find the key of a song if you would rather work it out by ear. Then count piano keys from the original key to your target, black notes included, and that number is your shift.
Job two: varispeed, the tape trick
Varispeed is the oldest pitch change there is: play the recording faster and everything gets higher, play it slower and everything gets lower. Tape machines and turntables do this mechanically. In software it means pitch and speed are locked together, so a few percent faster is also a little higher, and so on.
Sound quality is the reason to choose it. Varispeed does no time-stretching math at all, so there are no smearing artifacts to hear. Slowed material comes out darker and heavier, sped-up material brighter and lighter, and both sound natural in the way old tape manipulations sound natural, because that is exactly what they are. The tradeoff is that you cannot pick pitch and tempo separately: drop the key and the song gets longer by the same ratio, whether you wanted that or not.
This coupled sound is the whole aesthetic behind slowed edits. Our Slowed + Reverb Maker is built on this trick: slow the song and drop the pitch with a wheel, then export a WAV. The same move pointed the other way powers sped-up edits. One honest note if you plan to post these: publishing an edit of someone else's track generally needs permission from the rights holders (personal listening is a different matter).
Job three: slow it down without the chipmunk effect
This is the practice job. You are learning a solo, transcribing a vocal run, or trying to catch fast lyrics, and you need the song at 80 percent speed while every note stays at its real pitch. Otherwise everything you learn lands in the wrong key.
Time-stretching handles this: the tool changes the duration of the audio while holding the pitch steady. It is the most invasive of the three processes, because the algorithm has to add or remove time from a signal that was never meant to bend that way. Moderate slowdowns, roughly 75 to 90 percent of original speed, hold up well. Extreme ones get watery, and drum hits lose their snap.
For learning purposes that rarely matters. You are listening for which notes happen and when, not mastering a record. Set 85 percent, work through the hard bar, and step the speed back up as your hands catch up. The Pitch Changer also lets you set pitch and speed independently, so you can slow a song down and nudge it into an easier key in one pass.
If practice turns into recording covers of your own, our free desktop app handles the next step: record a take and convert your vocal to sound like a different singer.
What each method costs you in sound quality
If you remember one thing, remember this: varispeed is the cleanest, pitch shifting at a fixed speed is nearly clean for small moves, and time-stretching is the roughest of the three but usually good enough for its job.
Varispeed has no processing artifacts at all, but pitch and tempo are stuck together. Pitch shifting with the speed held steady is close to transparent up to about 3 semitones and warbles as the shift grows, especially on busy mixes. Time-stretching with the pitch held steady is fine at moderate speeds and smears at extremes. None of that should put you off. It just explains why a 7-semitone shift of a dense mix sounds strange, and why the old tape trick never does.
How to do it in your browser
Open the Pitch Changer, drop in your file, and pick the job: shift the key while keeping the speed, change the speed while keeping the pitch, or let the two move together for varispeed. Listen, adjust by a semitone here or a few percent there, and export when it sounds right.
Exports come out as WAV, which is lossless, so the tool adds no encoding loss of its own. One caveat stated plainly: if your source was an MP3, the export cannot restore anything the MP3 already threw away, so start from the best copy you have. And as with all our free tools, it runs in the browser, nothing uploaded, we never see the file.