The Dolby Atmos renderer accepts 128 inputs in total, and a full 7.1.2 bed eats ten of them before you place a single instrument. That leaves 118 objects. It sounds like more room than any song could ever need.
Most sessions never come close to the ceiling. That is exactly why the number gets ignored, and why plenty of mixes end up carrying dozens of object tracks that never move an inch.
Promotion is too easy. In a Logic spatial audio project, every new track starts as a bed track, and you convert it by clicking the output slot and picking the 3D Object Panner. Two seconds of clicking, no warning, no cost that shows up anywhere on screen. The choice matters more now that you can paste beat-locked panning written into object automation, because only a promoted track will hold it.
Stereo tracks are the quiet drain. A mono source routed to an object takes one slot. A stereo source takes two, so a session full of stereo synth stacks burns through the budget twice as fast as the track count suggests. Nothing in the mixer warns you as the count climbs, and the renderer stays quiet until you run out.
There is also a hard stop worth knowing before you plan the session. Surround tracks cannot be routed to an object at all. Mono and stereo can, surround cannot, so anything you already committed to a 5.1 or 7.1 stem stays in the bed whether you like it or not.
Beds also suit anything that does not need pinpoint placement. Wide pads, room mics, reverb returns, held strings. These are sources where a listener could not tell you within thirty degrees where the sound came from, and spending a slot to describe a position nobody perceives is a waste. The convention comes from film and games, where environmental ambience and background score live in the bed while the things you are meant to track move as objects.
The bed has its own ceiling. That format gives you one pair of overhead speakers, so you cannot separate ceiling front from ceiling back inside it. Anything that needs to live specifically above and behind the listener has earned a slot on that basis alone.
Some engineers working on music skip the bed entirely and run everything as objects. That works, and it sidesteps the height limitation, but it also means every element now carries position data whether it needs any or not.
Position is not the only thing that travels with an object. The metadata can also describe size, so a source reads as a tight point or a broad shape, and the renderer works out in real time what each speaker plays.
A static object parked at a speaker position is barely doing anything a bed channel could not do. You spent an input to describe a coordinate that never changes.
The value appears when the coordinate does change, or when a source matters enough that you want the renderer solving its placement separately on a soundbar, a 9.1.6 room, and a pair of earbuds. Lead vocal, yes. A hi-hat ticking between two ceiling positions on every eighth note, yes. The third layer of a pad that sits wide and still for four minutes, no.
Working out how many object slots a mix actually needs is a two-minute audit. Go through the session and mark every source whose position changes, plus every source you want placed precisely. Those get objects. Everything else goes to the bed, and the session gets easier to navigate as a side effect.
Most sessions never come close to the ceiling. That is exactly why the number gets ignored, and why plenty of mixes end up carrying dozens of object tracks that never move an inch.
Promotion is too easy. In a Logic spatial audio project, every new track starts as a bed track, and you convert it by clicking the output slot and picking the 3D Object Panner. Two seconds of clicking, no warning, no cost that shows up anywhere on screen. The choice matters more now that you can paste beat-locked panning written into object automation, because only a promoted track will hold it.
The renderer counts every channel you spend
A 7.1.2 bed is seven surround channels, one LFE channel, and two height channels. Every one of them counts separately toward the 128, which is where the 118 comes from. Run a narrower bed, and you free up inputs, though the arithmetic rarely rescues a session that promotes everything by reflex.Stereo tracks are the quiet drain. A mono source routed to an object takes one slot. A stereo source takes two, so a session full of stereo synth stacks burns through the budget twice as fast as the track count suggests. Nothing in the mixer warns you as the count climbs, and the renderer stays quiet until you run out.
There is also a hard stop worth knowing before you plan the session. Surround tracks cannot be routed to an object at all. Mono and stereo can, surround cannot, so anything you already committed to a 5.1 or 7.1 stem stays in the bed whether you like it or not.
Bed tracks handle some sources better than object tracks
Objects cannot feed the LFE channel. Only beds can. If you have an 808 or a sub layer you genuinely want sent to the subwoofer feed rather than distributed by the renderer, that material has to sit in a bed, and no amount of object automation changes it.Beds also suit anything that does not need pinpoint placement. Wide pads, room mics, reverb returns, held strings. These are sources where a listener could not tell you within thirty degrees where the sound came from, and spending a slot to describe a position nobody perceives is a waste. The convention comes from film and games, where environmental ambience and background score live in the bed while the things you are meant to track move as objects.
The bed has its own ceiling. That format gives you one pair of overhead speakers, so you cannot separate ceiling front from ceiling back inside it. Anything that needs to live specifically above and behind the listener has earned a slot on that basis alone.
Some engineers working on music skip the bed entirely and run everything as objects. That works, and it sidesteps the height limitation, but it also means every element now carries position data whether it needs any or not.
Movement is the strongest argument for a slot
The real difference sits in how the position data travels. Pan settings on an object track, static or automated, are sent to the Dolby Atmos plugin as separate metadata alongside the audio itself. The renderer then places that sound from the position data, adapting to whatever speaker layout the listener actually owns.Position is not the only thing that travels with an object. The metadata can also describe size, so a source reads as a tight point or a broad shape, and the renderer works out in real time what each speaker plays.
A static object parked at a speaker position is barely doing anything a bed channel could not do. You spent an input to describe a coordinate that never changes.
The value appears when the coordinate does change, or when a source matters enough that you want the renderer solving its placement separately on a soundbar, a 9.1.6 room, and a pair of earbuds. Lead vocal, yes. A hi-hat ticking between two ceiling positions on every eighth note, yes. The third layer of a pad that sits wide and still for four minutes, no.
Working out how many object slots a mix actually needs is a two-minute audit. Go through the session and mark every source whose position changes, plus every source you want placed precisely. Those get objects. Everything else goes to the bed, and the session gets easier to navigate as a side effect.
