James Cridland

Music streaming services, and cross-fading songs. And decent segues.

If you listen to pop music radio, you’ll notice that there’s never silence. Songs will merge into another quite nicely, and keep the energy going. There isn’t a silence between songs.

But if you listen to your Discover Weekly playlist on Spotify, or your Discover Mix on YouTube Music, you’ll notice that there is a silence between songs. It’s quite jarring, and if you’re trying to use music to keep the energy up, it’s even rather counter-productive.

The answer is not a cross-fade.

In Spotify, as one example, you’ll find a little feature put there by the devil called cross-fade. This is rarely the answer. Here’s why.

Talking Heads: Road to Nowhere starts with some acapella singing. “Well, we know where we’re going. But we don’t know where we’ve been".

If you set a cross-fade, then Spotify will fade out the previous song, while fading up this song. Set a five second cross-fade, and it’ll cut off the first few bars of the track, and make it start “… we’re going”. And that always sounds awful. It can’t ever sound good.

The Power Station’s Some Like It Hot ends abruptly, with a snatched lyric of “Some like it Hot.” If you fade this early, you ruin this song. You need to wait until the precise tenth of a second that the song has finished and then start the next one.

That’s quite different from The Beatles: Let It Be which ends with a long sustained note. The song has finished. The last note is held. If you wait until the end of this song, all the energy has gone, because you’ve been listening to five seconds of a final note slowly reverberating away.

And that’s a very different ending to Rick Astley’s Never Gonna Give You Up, which fades, like almost every Stock Aitken and Waterman creation does.

And Nik Kershaw’s The Riddle actually fades in.

To mindlessly throw a cross-fade on these incredibly different songs is clearly the wrong thing to do.

The answer is a decent segue

No, you don’t need to beat match or do anything clever like that. They don’t beat match on the radio (often). They just know when to start the next song.

Any radio DJ will listen to the start of the next song on prefade (i.e. just for themselves, nobody else) so they know how it goes. They’ll also have the intro time written down on a piece of paper in front of them probably. And the paper also says whether the song ends or fades.

This is not beyond the wit of software. Even software I had in 2001 was capable of a decent segue (step forward, OTSJuke).

On a basic level, it should work out when a song goes below a certain level at the end. No AI required. Then, the software knows when to start the next song. It’s that simple. (You might start fading out the previous song once you’ve started the next one: it sounds cleaner if you do.)

It can be more complicated than that: but even setting a trigger point like this would make music services work much better. You only have to set that trigger point once.

I wonder why they don’t?