Where two songs meet
How the Orage host talks over the edges of songs, and how one song becomes the next.
Every few songs, Orage has something to tell you. A review came in. A build failed. Someone needs an answer. It does not stop the music to say it. It waits for the end of the song, and says it there, the way a radio host does: over the last bars of one record and the first bars of the next.
Listen to one first. Then I will tell you how it is made. This one played on October the eighth, in the morning. Signatune, by DJ Mehdi, is ending. Justice is next, with Genesis. The host has two things to say. Listen for the music dipping under each sentence, for the breath between them, and for the last word ending just before the drop.
Transition 9MA-A4Zx as Orage saved it on October 8, 2026: the real songs and the real host line, mixed live on this page by the app's own level rules. The music fades at the end so the story can go on. In Orage it plays on.
How it was made
That was about twenty seconds, and almost everything Orage knows about sound went into it. It starts with a plan, made before anything plays. The planner looks at the two songs: the one that is ending, and the one about to start.
It knows the shape of each song: its beats, its bars, how loud it is at every twentieth of a second, and where the singer comes in. From that it chooses the landing: the moment in the next song where the talking must be over. The best landing is a drop, a jump of five decibels on a beat, the music rising out of something. Genesis has one, thirty-nine seconds in. If there is no drop, the host finishes a beat before the singer. If there is no singer, it finishes before the biggest change of section.
Meanwhile the host has written its line, four sentences, and ElevenLabs has said them, returning the time of every character, so the planner knows how long each sentence lasts, to the millisecond. Then it walks backwards from the landing. The sentences are placed on beats, from the last to the first, with at least half a second of silence between them, and the next song plays four beats alone before the host goes on. Whatever does not fit in the intro goes over the end of the song that is ending, in its last thirty seconds, on a breakdown if it has one, never earlier, so the song is not cut short. Here, that is the first sentence.
The sound under the voice
While the host talks, the music does not stop. It dips. Nine decibels down in a quarter of a second, just before the first word, and back up in half a second after the last one. Between two sentences the music comes back for a breath, then dips again.
The voice itself is treated like a radio voice. Host clips come about fifteen decibels quieter than a mastered song. A gentle compressor evens the speech, then the line is raised to sit one decibel under the louder of the two songs, never louder than minus nine LUFS, with a limiter holding its peaks. Over music dipped by nine decibels, that keeps the voice eight decibels clear.
The ending song fades out over two seconds once the next one comes in. The next one fades in over a second and a half, and stays down, under the voice, until the landing. At the landing, the music is full.
- 1 DJ Mehdi gives way to two decisions.
- 2 The Daytona preview and checks are good, but you need to update the production key before release.
- 3 Lou needs your consent for an isolated install to finish checks without changing project files.
- 4 Now, Justice with Genesis.
The plan of transition 9MA-A4Zx, drawn from the saved record: the two songs over the same forty-nine seconds you just heard, the levels measured on the songs, the host's sentences where the planner put them, and the dips and fades from the app's own rules.
Why it matters
A host that interrupts is a notification with a voice. A host that waits for the edge of a song, dips the music, and lands on the drop is something else: a presence. It respects the music, so you can trust it with your attention. Most of what you heard was silence, placed on purpose. That is the talent Orage is built around: not the voice itself, but when it speaks, and what it leaves alone.