Ngl, I’m always whining about not having enough time and having way too many things I want to do. I wish I had infinite time. So of course I went and bought the Vocaloid (in a broad sense) software I’d been eyeing for a while, Synthesizer V 2 AI Kasane Teto, and it ate even more of my time.
Kasane Teto Original Song: Milk and Sunbeams 🎵
First, here are the results. I’ll explain how it happened later.
YouTube ▶️
Nico Nico Douga 📺
If it’s a Vocaloid song, it’s gotta be on Nico Douga too, right? So I posted it there as well. It’s been years since I last posted anything on Nico Douga, so I even made a brand-new account for it.
Okay, I’m Obsessed with Kasane Teto 🎤
I already liked Vocaloid songs sung by Kasane Teto, but the other day I bought VOICEPEAK Kasane Teto, and my brain was fried by the thrill of making Teto-san sing whatever I wanted.
I’d wanted Miku’s Vocaloid software for ages, but I kept holding back because I knew for sure I’d buy it, feel satisfied, and barely use it. But Teto-san? Couldn’t resist, man.
What’s great about Teto-san is that her energetic voice is nice, of course, but I absolutely love her breathy low voice, the vibrato that kicks in, and her relaxed singing style.
I already prefer voices that sing like that, but there’s just something irresistible about how it sounds like she’s right next to your ear.
So, I decided that the first song I make would be exactly that kind of song.
Lyrics, Composition, Arrangement, Mixing… All Me 🎹
I had never composed a song before. Not once.
I looked into what needs to be done to finish a song, and wow, there is a LOT to do.
Which makes sense. Every song we casually listen to is a huge pile of sounds, and every one of those sounds was chosen by someone: the person who made the song. Obvious, but it really hit me.
How It Actually Went Down 🛠️
Whether the lyrics or the melody comes first seems to depend on the person, but I felt like I couldn’t move forward without grasping the vibe. With just the theme of “relaxed singing” and a “stylish feel,” I decided I wanted her to sing loosely over a Bossa Nova/City Pop style track.
I had a vague idea that the order goes Drums → Chords → Bass → Melody, so I proceeded with the composition in that order this time.
The structure went roughly like this: Intro → A-Melody → B-Melody → Chorus → Modified B-Melody → Final Chorus → Outro.
The drums were pretty simple. Kick, snare, and closed hi-hat hitting on the offbeat to give it a Bossa Nova feel. I just roughly programmed that all the way to the end.
Next I tried to come up with a chord progression, but I had no idea.
So I ended up binge-watching YouTube videos on chord progressions and punching them one by one into my DAW (Digital Audio Workstation: software for making music on a computer) to compare how they sounded.
I went with vibes, since I don’t know anything about music theory.
The base is the so-called “royal road” progression (王道進行, IV→V→iii→vi) that I’d heard about, but I tweaked things here and there, picking chords that sounded good to me to build out the sections.
Programming MIDI by hand was such a pain that I ended up making a separate tool halfway through to deal with it.
I might release this someday, but it’s a rough temporary tool that looks pretty lame, so I’m a bit too embarrassed to publish it as is.
Functionally, of course, it meets my requirements. It lets me copy template progressions, or build chords by combining root notes myself.
I designed it so you can build chords for 16 measures.
You can output the finished progression as MIDI, so making the chord progression with this and importing it into a DAW saves a ton of time clicking around.
Just hearing all the different chords is kind of fun even if you just hit buttons randomly.
Next was the bass. I made sure not to play the root note (the lowest note) of the chords on the piano, and had the bass handle that instead.
Then I just clicked around using those notes until it sounded good.
I actually really like the sound of the bass. I love how it supports the band.
Then I moved on to the melody, and this is where I got stuck.
People on YouTube say you can just hum a melody while listening to the chord progression, note down whatever comes out, piece it together, and you’re mostly good.
Nothing came out.
I had no choice but to think of the lyrics first.
I had a theme for the lyrics in my head, and since I want to leave the interpretation to the listeners, I won’t go into detail here, but I managed to write them without too much trouble.
There was one entire set of lyrics I scrapped, but I might use it someday, so I’ll save it for later.
Based on the finished lyrics, I divided them into parts and created melodies so they would fit into each measure.
It took forever. Seriously, the struggle is real.
Then, it was finally Teto-san’s turn, and man, this was fun!
Teto-san actually sings for you (obviously).
I got pretty absorbed in programming her vocals. Of course, since it was my first time, I’m sure there were a lot of inefficient steps.
But making three parts (main melody, harmony, and humming) and listening to a track full of Teto was just bliss for my ears.
It took a lot of time, but I managed to finish a full song. It’s a short track at 2 minutes and 30 seconds, but I’m satisfied.
Mixing: The Mysterious Black Box 🎛️
Mixing. As I understand it, it’s the process of taking the finished song, processing each instrument nicely, and balancing the overall track.
I might be wrong.
Anyway, I had heard the word “equalizer” before but didn’t know what it actually was. I studied its effects and adjusted things so the instruments wouldn’t clash at the same frequencies: cutting the lows here, cutting the highs there, or slightly boosting the mids.
I positioned each instrument front, back, left, and right around the vocal so the sounds wouldn’t step on each other, while making sure they didn’t sound too disconnected either.
I applied a reverb effect to adjust the echo, and tweaked the overall loudness so it matched commercially available music.
I definitely can’t do one of those “cool” mixes, so I just stuck to the basic shaping needed to make it function as a song.
Even so, there’s a strong impression that the sound changed drastically between the unprocessed and processed versions.
Mixing is amazing.
Again, this involved dealing with words I didn’t know 90% of the time, watching videos, applying it to my song, and adjusting over and over.
Sometimes I’d go back and tweak the tuning (Teto-san’s voice) too.
Because the sound changed so noticeably, I feel like this process accounts for about 30% of the whole song creation effort.
Pardon my amateur opinion, though.
I listened to it in five different environments: PC + headphones, PC + cheap speakers, phone + earphones, phone + speakers, and car speakers. I realized how incredible headphones are. Earphones too.
It really made me understand how differently the exact same sound can be heard depending on the environment.
I exported six versions in total (the final one is named “Deploy 6”).
Done (For Now) ✅
Anyway, it’s done. Since I was the one making it, if I say it’s done, it’s done.
But this project couldn’t end just there.
Because if you make something, you want to publish it, right?
So, I decided to go straight into making an MV.
However, making a highly dynamic, moving MV is beyond my skill level.
Instead, I opted to make something like a lyric MV (a single illustration plus lyrics), which you often see with Vocaloid songs.
Drawing Teto-san 🎨
I wanted to animate the drawing at least a little bit, so as a bare minimum, I drew three variations of the eyes.
The other one is just her eyes closed with the eyebrows slightly raised, so I’ll skip showing it here.
Actually, this Teto-san is the cutest I have ever drawn.
I managed to give her a slightly mature vibe, and I want to praise myself for getting better at drawing. (Since nobody else will praise me.)
I was really happy to get around 40 likes on X (Twitter) too.
The production time was just over 11 hours.
Lately, the time it takes me to finish one piece has settled around this mark.
It looks like this.
Obon week (Japan’s mid-August holiday week) stands out, but even that was only 15 hours in total.
I wonder how this compares to people who draw seriously…
Still, I’m pretty satisfied with how this Teto-san turned out, so I want to keep improving at drawing her.
Making the Video 🎬
Video creation. As always, I used AviUtil2.
It’s honestly godly that such a multifunctional tool is free.
Since my video editing skills are still low, it takes me an agonizingly long time, though.
I roughly know my way around it, so I steadily put together a lyric MV.
Even though it’s a lyric MV, since the song has a relaxed, chill vibe, I kept the lyrics from zipping around and mostly just used light fade effects.
Neither the song nor the video has any elements that would make it go viral, but I hope people can listen to it leisurely for a long time.
The Wrap-Up 📝
To sum up.
As a complete music amateur, I bought Synthesizer V 2 AI Kasane Teto and did everything (writing lyrics, composing the song, arranging, mixing, drawing the illustration, and editing the video) and uploaded it.
A true one-person production.
No surprise, the hardest part was the completely new experience of composing a song.
It took so much time, and I managed to finish it without ever really knowing the “right” answer. Just when I thought I was done, I got bogged down in the mysterious process of mixing. As usual, the illustration took over 10 hours, and finally, editing the video for a 2.5-minute song took me about 3 days.
It might be a sickness at this point that whenever I make something, I feel the need to publish it. But going through the whole process from composing to publishing this time, I realized that while it’s fun, continuing this consistently is incredibly tough, and I gained a deep respect for Vocaloid producers.
Now I know firsthand what this kind of work costs.
I posted it on video sharing sites, but I don’t really expect my first song to become popular. Still, since I went through the effort to make it, I’d love for people to listen to it for a long time to come.
I actually have an idea for a second song, but it seems I’ll be busy for a while, so I probably won’t be able to dive into composition work anytime soon.
Next time I start composing, I think it might be fun to post screenshots and progress updates on X (Twitter) as I go.
Bluesky Discussion
View on Blueskyまだ返信はありません。Bluesky で最初にコメントしましょう！
Comment