How playback is scheduled¶

What actually happens when you press Play — how clips, loops, and the playhead turn a static arrangement into sound over time.

One transport, one clock¶

There is exactly one transport: one tempo, one playhead, one clock the whole song follows. The playhead's position is measured in beats (a beat is a quarter note). Everything below is that clock deciding what to sound now.

Playback happens block by block¶

The audio engine doesn't play the song note by note — it works in small blocks (a few hundred samples at a time). For each block, it asks a simple question: which notes and audio fall between where the playhead is and where it'll be at the end of this block? Anything in that window is emitted at the right sample offset inside the block.

This is why timing is tight and why the whole thing scales: the engine only ever looks at the sliver of time it's about to play, not the entire song.

Clips are anchored, and they loop¶

A clip sits at a fixed position on its track's timeline (startBeat). Its content has a length, and by default a clip loops to fill however long it's placed:

Drop a one-bar idea across a four-bar span and it repeats four times. You don't copy it — the clip's content just wraps every bar as the playhead crosses it.
Turn looping off and the content plays once, then the clip is silent for the rest of its span.

So a clip's placement (where and how long) is separate from its content (the notes/audio inside), and looping bridges the two.

The loop region wraps the playhead¶

That's a clip looping. Separately, the transport loop — the region you set with the Loop button — wraps the playhead itself: when it reaches the end of the loop, it jumps back to the start and keeps going. Everything inside plays again because the clock literally went back in time. This is what lets you record pass after pass over the same bars (see loop-record and takes).

Session clips play alongside, live¶

A launched session clip is scheduled differently: it has no timeline position, so it loops from the moment you launched it, in its own running beat count. While it plays it overrides that track's arrangement playback; stop it and the arrangement takes back over.

Session launching is a live decision — it depends on what you fired and when — which is exactly why it's kept off the render path (below).

Automation and modulation move per block¶

Time-varying values ride the same clock. Each block, the engine reads any automation lane and modulation LFO at the current playhead position and applies the value for that block. So a drawn fade or an LFO wobble is evaluated continuously as the playhead moves — no separate pass.

Offline render is the same clock, deterministic¶

A render runs this exact block-by-block scheduling, but freewheeling — as fast as the CPU allows instead of in real time — over the arrangement. Because it reads the same fixed clips, tempo map, and automation, it produces the same file every time: no live session launches, no real-time jitter, no dropouts. That determinism is the reason renders come only from the arrangement, never from live session playback — see Arrangement vs. Session.