How I Did It
Uses samples from:
Tools I Used
Acer Aspire One netbook, Beyerdynamic DT990 headphones, Audacity with GSnap VST plugin, jeskola buzz tracker. Mp3 encoded with LAME and Razorlame GUI.
Process
It was quite a challenging process for me. It took some extra time and I learned new things as well. Here's what happened, roughly.
It all began by cutting the original vocal recording into short samples (a few seconds or one to four lines of lyrics). It's the easiest way to work when using a tracker as the music editor.
When using those samples, I noticed the vocals are out of tune. I decided to look for free pitch corrector, and I ended up using GSnap (VST). As I used jeskola buzz tracker for creating the music, and it is not a normal audio/midi sequencer, it was clear I wanted to first correct the vocals, and then use the corrected vocal samples in buzz, just like my original idea was.
Soon I was learning how to use an autotune effect. It took some time until I understoood all the parameters and how they affect the sound - and GSnap doesn't even deal with formants! By being unable to control the GSnap via MIDI, some vocal lines were difficult, some impossible to correct, unless going for smaller segments of audio. Luckily, the GSnap worked fine in Audacity! I worked a line or sometimes a word at a time, manually choosing the correct notes used in the selection in question. Doing multiple GSnap operations over a line causes pops or discontinuities, which needed to be fixed by careful editing. Moreover, the GSnap sometimes missed a period of the waveform (the female vocal normally consists of two different periods alternating), causing a lower midrange "pop". Then, locating the pop and carefully deleting or copying a single period to make the waveform periods appear in pairs again.
Some lines were too complex to fix with GSnap, but their actual problem was shifted tuning only, so I decided to use the Change Pitch effect in Audacity for those lines. That caused a problem: GSnap output has reduced amplitude and high end (treble), so the fragments were incompatible. I needed to "correct" the amplitude of the GSnapped fragments and EQ the Pitched fragments.
At this point I should tell this: don't get me wrong, I have heard from other tracks that Silvia O sings perfectly well. The reason there was correction needed is likely because the original vocals had been sung without any background music playing, so the tuning was just a bit off, but for me tuning my gear was not an option.
I had been testing my melodies and chords a bit before the vocals had been fixed, but I got stuck with the minor scale. After I had fixed the vocals, one day I just got a new idea for the melody, which requested for major scale chords, and it all felt better - the project was saved again!
Then it finally went along my normal "lines" of production: progressing slowly in fragments, limited a lot by the available time. Eventually I got one main loop based live version recorded, and had been enjoying the sound for quite some time, until one day I felt like I should try making the biggest step: polish the fragments of tracks and the mix and arrange the whole thing as a complete song. For me it's usually the most demanding task. I can get some good ideas, sounds and melodies, but how to stretch and modify them to fill an entire song - that's the question. I usually don't know half of the result to be, until it's done. The music is the guide :)
Other Notes
Creating this remix was a really nice experience, because I learned a lot about pitch correction. Most shockingly, I learned that it is really possible to do corrections that are somewhat impossible to notice!
What also makes this so special, is that this is my first contribution to ccMixter, and is based on the first ccMixter original I have worked with.
I also want to extract and upload the vocals I corrected, as my time permits, for others to have better chance on creating new remixes of Un solo corazon!
This site uses