Transcription
You can write the best lyrics in the world, but if your AI singer sounds the same from start to finish, the song falls flat. That's the thing nobody tells you about Sununo.
By default, the AI singer hits every note at the same level of physical effort from the first second to the last. The pitch changes, the melody changes, but the intensity of the voice that stays flat. And that's a problem because real songs aren't sung at one level.
Okay, think about your favorite emotional track. The singer probably starts soft, maybe almost whispering. Then they ease into a more grounded voice for the verse. By the time the chorus hits, they're pushing their voice to the limit. Full chest, full power, full body. That journey is what makes a song move you. The voice itself physically changes as the emotion of the song changes.
Today, I'm going to show you how to recreate that in Sunno using something called vocal register tags. These aren't style tags. They're not genre tags. They're a different category. They tell the AI singer how to physically use their voice in each section of your song. Let's break it down.
In real world vocal performance, a singer can place their voice in different registers. A register is basically how the singer is producing the sound, where the voice is sitting in the body, how much air is moving, how much projection they're using. Sunno has been trained on enough vocal performances that it actually understands this language. When you tag a section with a specific register, the model adjusts the physical effort of the delivery to match.
There are eight registers that Sunno responds to. Four of them are the ones you'll naturally use the most. The other four are situational, useful, but not the core toolkit. Let's start with the core four.
Falsetto. This is that light, airy head voice sound. The voice sits high and feels almost weightless. Think of those breathy, dreamy moments in a song where the vocal floats above the track. Falsetto is your tool for ethereal intros, vulnerable bridges, or any moment that needs to feel delicate.
Chest voice. This is the grounded, full-bodied, speaking, singing voice. It's where most conversational verses live. The voice sits in the chest. The tone is warm and present. Chest voice is your default for storytelling sections. Versus where you want clarity and confidence, not high drama.
Mixed voice. This is the bridge between chest voice and the higher registers. It blends warmth with reach, which makes it perfect for buildup moments. Pre choruses live here. Anywhere you want the energy to climb without fully exploding yet.
Belted. This is the big one. Belting is when the singer pushes their voice to the absolute limit while still keeping it controlled. It's the explosive "I'm going for it" sound and themic choruses, climactic moments. Belted is what makes a hook feel huge.
Those are the four registers that will cover 90% of what you'll ever need to do in Sunno. The other four are worth knowing about briefly. Head voice, high and pure, similar to falsetto, but with more body, whispered, breathy, and intimate. Use this carefully. It's the hardest register to land consistently. Spoken word. Exactly what it sounds like. Used for talking sections, intros, or interludes. Raspy, gritty, rough texture. Good for emotional grit or rock adjacent moments.
Now, let's actually use these. For the first demo, I'm going to do something dramatic, a falsetto to belted contrast. This is the biggest possible dynamic shift you can make with registers. The voice goes from light and floating to full power and projection. The contrast is huge, which is exactly what we want to hear. The genre is dark pop, emo pop. The lyrics tell a story of internal collapse turning into emotional release. Notice what the tags are doing. The verse is marked falsetto, fragile. That tells to keep the voice light, airy, almost breaking. Then the chorus is marked belted, explosive. And we even wrote the lyrics in capital letters to emphasize the emotional shift. Let's generate.
[music] >> Holding on to pieces of fading light. Maybe I was never [music] meant to feel this small. Maybe [singing and music] I was meant to lose it all. BUT I'M STILL standing in the wreckage, screaming at [music] the sky. Every [music] >> That's the contrast. The verse feels like the singer can barely hold it together. Then the chorus hits and the voice physically erupts. Same singer, same song, but the body behind the voice has completely changed. That's vocal placement at work.
Okay, let's try a different combination. For the second demo, I want to show that this trick isn't just about going from soft to loud. It can also work the other way. Starting grounded and then drifting into something lighter. This time we're doing chest voice in the verse, then dropping into falsetto. For the hook genre, we have smooth soul R&B. Let's hear it.
>> I was walking through the city in [music] the heat of June. Thinking about the way you said you'd be here soon. [music] Every street corner looked just like you. Every quiet moment had your [music and singing] name running through. You're the kind of love that lingers [singing] [music] like a melody that lingers [music] on. Ooh, I can feel you in the silence. [music] Even when I know that you are gone. [music] >> Hear how the verse at the beginning sits in the chest. It's grounded, conversational, like the singer is telling you the story face to face. Then the chorus lifts. The voice goes light and airy. It feels like a memory floating through the song. That's the same singer in two completely different physical states. And we directed both of them.
Now, let's talk about where this trick breaks. First, fast tempos. If your song is moving at high BPM, the singer doesn't have time to actually shift their vocal placement. They'll just default to whatever's easiest and rush through. Second, dense lyrics. If you've crammed 30 syllables into a short line, the singer is too busy pronouncing words to physically shift how they're producing them. Give the singer space. Fewer syllables, longer notes, and the register shifts will land. Third, overuse. If you switch registers every two lines, the singer sounds erratic, glitchy, like they can't decide what they are. Stick to two or three register changes across the whole song. One per major section is plenty. Fourth, genre conflict. If your style prompt is soft lowfi piano, but you tag the chorus as belted maximum intensity, you're asking for two opposite things. The AI will get confused and either ignore the tag or produce a messy take.
So, how do you get this right consistently? Apply changes between sections, not mid verse. Your verse should feel like one performance state, and the chorus should feel like a different one. That contrast is the whole point. Midsection register changes confuse the model. Build a performance arc. Most great songs start lower in intensity and build upward. Soft intro, grounded verse, climbing pre chorus, explosive chorus. The register choices should mirror that emotional climb. Match the register to the actual feeling of the lyric. Vulnerable line, falsetto, confident statement, chest voice, climbing realization, mixed voice, triumphant release, belted. Let the meaning guide the register choice and keep it simple. The best vocal arrangements use a handful of register shifts, not a register for every line. Restraint is what makes the shifts hit when they happen.
Now, let's put everything together into a full performance arc. This is where this trick really shines. For the final mega prompt, we're doing a modern country ballad. Country ballads are perfect for this technique because they tell emotional stories that build, and country production naturally leave space for dynamic vocal shifts. We're going to use all four core registers in one song, falsetto, chest voice, mixed voice, and belted to walk the song through a complete emotional journey. Notice the arc. The intro starts in falsetto, fragile, almost a thought rather than a statement. The verse drops into chest voice, grounded storytelling like she's telling you what happened. The pre chorus shifts to mixed voice as the emotion starts to climb. Then the chorus explodes in a full belt and the final chorus stays in that belt at maximum intensity. No holding back. Same singer, four different physical states, one emotional journey from start to finish. Let's generate the porch [music] lights. I drove past [music] the old church where we used to talk. counted [singing] every mile from the truck to the dock. The radio [music and singing] was playing that song you loved, the one about leaving [music] and the things we lost. And I [music] felt it rising in my chest [singing] again. That same old feeling I can't pretend. So I'm driving through the night with the windows open wide. Every memory I've been holding. Coming [music] along tonight. I'm driving through the night. Nothing left to hide. [music] Every part of who I've been is wide [music] awake tonight.
That is just terrific, right? That's the difference. The vocal isn't flat anymore. It moves. It breathes. It builds. The singer isn't just hitting notes. She's living through the song. And that's the point of this trick. That's how you turn a vocal track into a vocal performance. You're not just telling Sunno what notes to hit. You're directing how the singer should physically deliver every section. Where to hold back, where to build, where to let go completely. You're shaping the emotional architecture of the entire vocal.
Try this on your next track. Pick the most vulnerable moment and tag it falsetto. Pick the most grounded section and tag it chest voice. Pick the biggest emotional release and tag it belted. Build the arc. You'll hear the difference immediately.
If this breakdown helped, subscribe. There's a lot more coming on the science of getting professional results out of AI music. Keep crafting. See you in the next one.