[# WRITTEN PRAYERS — "Billie Eilish"] [### single female vocal · one male intrusion] --- [## ACT I — GIVE HER ROOM] **[Intro — whispered]** God... before the lights, before the numbers, before everybody knew her name— there was just a girl trying to become herself. Before the world decided who she was. Give her room. **[Verse 1]** Give her somewhere nobody needs anything. Somewhere tired can simply be tired. Somewhere she can laugh, change her mind, walk alone, and not be watched. Let her remember before the world started keeping score. **[Transition]** The numbers keep climbing. One becomes millions. A bedroom becomes a stage. A voice becomes something strangers think belongs to them. And still— she is one person. One heartbeat. One life. **[Pre-Chorus — obsessive, quickening]** What's next? Another song. What's next? Another stage. What's next? Another number. Teach her to stop. **[Drop]** What's here? This breath. This room. This heartbeat. **[Chorus]** Let her be here. Not famous. Not watched. Just here. Let this moment be enough. This breath. This heartbeat. This moment. --- [## ACT II — SOMEBODY ANSWERS] **[Verse 2]** Let her miss somebody without making the missing disappear. Let her hurt without rushing toward healing. Sometimes we don't need another lesson. Sometimes we need permission to feel. **[Bridge — quiet, direct]** When everybody's watching but nobody sees you— when your pain becomes a chorus and strangers know the words— do you still get to keep it? God, who's taking it from her? *(silence — two bars)* **[Male Rap — the only other voice in the song. Uninvited.]** Me. I watched her on a screen in a room she will never enter. I paused her at the worst hour of her life and I took notes. I wrote down what she was afraid of because it sounded like something I was afraid of. Everybody wants another piece of the person they think they know— and I said *everybody* like I was standing somewhere else. I took a girl's question and I made it my song. I put my name on it. I asked for room for her from inside the crowd that took the room away. So don't let me off. **[Female — after a beat]** I wasn't going to. You did. And you meant it. Both things are true. The taking and the meaning it. That's the whole problem with being a person. --- [## ACT III — GIVE ALL OF US ROOM] **[Verse 3]** Then let it be a real prayer. Not for the girl on the screen. For the girl. And the room. And every stranger who found a piece of themselves inside somebody else's song. Lonely people. Hopeful people. She didn't save them. She only said: *you are not alone.* Sometimes love is sitting beside the wound instead of asking it to sing. **[Breakdown — sparse]** When she says, I'm afraid— don't tell her she isn't. When she says, I'm tired— don't tell her to keep going. When she says, I miss them— don't teach her how to stop missing. Sit with her. Stay with her. Let silence say what words can't. **[Final Chorus — larger, still restrained]** Let her sing. Let her stop. Let her grieve what she lost. Let her go where we can't follow. Let her become. Let her be here. Not famous. Not important. Not watched. Just here. And let us be here too— not taking. Just here. **[Outro — whispered]** What's here? This breath. This heartbeat. This moment. Let it hurt. Let it matter. Let it pass. And while it is here— let me be here too. Amen. --- [## Production notes] [**The male section is the only voice change in the song.** One switch, one return. That is roughly the most a Suno generation will hold reliably across five minutes.] [**Give it a delivery contrast, not just a gender tag.** The model tracks *spoken / rapped / sung* far more reliably than it tracks male / female. If the intrusion is rapped and everything around it is sung, the form carries the separation and the tag becomes a backup rather than the mechanism.] [**The silence before "Me." is load-bearing.** Two bars of nothing after the question. Let the answer arrive into an empty room.] [**Watch the low end.** In the earlier duet, the male voice pushed 80–250 Hz from 26% to 35% of total energy in a mix already sitting 57% below 250 Hz. Carve for the rap section or it will crowd out the negative space that makes the rest of the song work] [**If the model drifts anyway**, generate Act I + II as one pass and Act III as a second, then join. Two shorter passes hold their voice assignments far better than one long one.]
