So I tease about how if a vocalist uses pitch correction then the vocalist must not be perfect, but naturally I too use pitch correction because it just sounds so damn nice sometimes. There is a point where vocals are no longer just vocals and they become a synth of it's own, but there's a sweet spot in between that sounds especially cool. When I listen to songs by Owl City and Fox Stevenson, I know the vocals are clearly corrected, but to what degree and in what way isn't immediately obvious because it still sounds fairly natural. I managed to figure out what I believe is the most probable method they use to get this sound.
For this example, I'm recreating the vocals from a part of the Fox Stevenson song called Endless.
1. Record the vocals at a lower BPM.
It's much easier to enunciate and keep proper pitch when singing to a lower BPM. The trick is finding a good one to use, because anything too close to the original will sound too natural, but drop the BPM too much and it'll sound too altered when we reach step 2. For this track, I recorded my voice at 80 BPM, which is lower but not too much lower than the original 105 BPM the project was at. Try to "speak slower" but keep your words as clear as possible. Avoid creating vibrato, since this may sound a little weird when we reach step 2. Also, I'd recommend normalizing the sample since this helps with step 3.
2. Speed up the vocals using time stretching to the song's actual BPM.
There's a way to do this in Edison, but I didn't care to find out. What I did was I put my vocal sample into the channel rack and then changed the time stretch mode to E3 generic. After that, I typed in the sample's BPM (although since I had recorded the vocals directly in Fl Studio, the BPM were embedded automagically into the sample), and then set my project back to it's original BPM. I then recorded this stretched version again using Edison so I could use it with the next step...
3. Clean up the vocal sample using Newtone.
This part was tricky. Here's what I did:
- I set "Center" to 100%, "Variation" to about 35%, and left "Trans" at 100%.
- I then went through and made sure all of my notes were correctly detected. For the ones that were off, I had to manually move them to the correct note.
- In some cases, two notes that were meant to be separate (and probably of different pitches) were detected as a single note. For those, I had to manually split the note.
- In some cases, a single note was automatically split into multiple parts. This is fine - there's no need to merge the notes, and this can actually be useful since Newtone often splits notes between vowels and consonants if the transition is strong enough. If the two notes aren't the same pitch, however, I'd recommend moving the one that's on the wrong note to the correct note but only if you can do so without obvious artifacts! The point is to keep it sounding natural, so do this wisely.
- In some rare cases, Newtone would split a note in a bad spot. Just merge the two notes together, center it, manually split it in the correct spot, and center the notes again.
- I then went through and carefully shifted boundaries between notes to get the timing of my notes and my consonants as synchronous as possible with the intended timing. The best way I found to do this was to set Newtone to be slaved to the host. I could then preview what the corrected vocal sample would sound like when played with the project. I also had the metronome turned on, which was the most important part since it helped me most when trying to get the timing right.
- Whenever I needed to split notes, I aimed for points between consonants, vowels, and silence (or breaths). This was often difficult to do with snapping on, so I'd often have to go and turn snapping to "Off" before trying to cut a note. After that I'd turn snapping back on since it can be useful.
- Whenever I tried to move boundaries between notes, I'd sometimes hold down the 'alt' key so they didn't snap. In some cases, this sounds more natural, since our voices aren't naturally quantized.
- Always, always, ALWAYS listen to each tweak before moving on. And don't be afraid to Undo changes
- Don't treat these directions as strict instructions! This is just my advice, after all. Using Newtone to produce the best possible sounds requires practice.
I'm still trying to figure out if there's a way to do this... I hate working on correcting a vocal sample and not being able to save my changes so I can edit them later. However, this step is optional.
*. Now you could just export this corrected sample and be done, but I took it a few steps further for the sake of recreating the original sound...
5. Send the notes from Newtone into a pattern for later use.
Just create a new 3xOsc in the channel rack and then create a new pattern. You'll be using this later.
6. Drop the now corrected vocal sample into the project.
Get it lined up right, get it at the right volume, and get it going through the correct mixer track. Make sure to mute or remove other versions of the vocal sample from the playlist. Also, if you still have Newtone opened, you'll want to disable "Slave playback to host" now.
I also noticed that there's another vocal sample in the background that sounds like the original but dropped down an octave. Use Newtone for this, and make it much quieter.
7. Setup a Vocoder!
Since I used Vocodex, this will require at least 3 other mixer tracks - one for the Vocoder, one for the Carrier (synth), and one for the Modulator (voice). Earlier on, I'd setup two other mixer tracks for my Vocal In and my Vocal Out, which allowed me to send the unaltered Vocal In directly to the Modulator mixer track and to the Vocal Out. I made sure the Carrier and Modulator mixer tracks were routed only to the Vocoder mixer track, placed Vocodex into the Vocoder mixer track, and set it up so it was accepting input from the Carrier and Modulator mixer tracks respectively. Getting the vocoder to sound right wasn't easy (I'm still not quite satisfied with the way mine turned out :/), so you'll need to look elsewhere for how to use that.
Oh! Remember that 3xOsc from step 5? Yeah, you'll need to route it through the Carrier mixer track now, since it's going to generate our carrier signal. I'd recommend using a saw wave shape
8. Setup your notes for your Vocoder.
So the pattern we created in step 5 will now need to be aligned so it plays in time with your vocal sample. You'll probably want to go into the pattern and make sure all of the notes have their velocities set to 80% as well, which can be easily done using the "Scale Level" tool in the piano roll. In addition to that pattern, we'll need to make another which contains the chords to that part of the song. That pattern will also need to be aligned. Lastly, you might need to change the octave of your carrier signal generator so it plays at the correct octave.
Unless I forgot something, that's about it!
Please play around with the project and feel free to post your ideas and your own edits below. (I'm especially curious if we can get the vocoder bit to sound right.
)- Mr. Raubana