Multichannel phasing and object export, free, runs in your browser.
One sound is played many times at once, and every copy runs at a slightly different speed. They start together. Over the next few minutes they slide apart, because a copy running 0.2% faster gains about a fifth of a second every minute. What began as one voice becomes a cloud, and if you leave it long enough the cloud closes back into a single point.
That much you could do in stereo. What makes this useful is that every voice is kept on its own channel, so instead of a wide stereo smear you get a specific number of separate signals that you can place anywhere in a room. Twelve voices become twelve Atmos objects, each drifting at its own rate, each somewhere different overhead or around you.
Making a sound
Pattern is the quickest way in. Pick Pulse, Mallet or Clap, set a step pattern and a tempo, and press play. Every voice runs the same pattern at its own speed.
Sample uses your own audio. Drop in a loop, a vocal phrase, a room recording, anything. Stereo and multichannel files are folded to mono first, since the app needs one signal to copy. You choose how that fold happens: sum both sides, take one side only, or take the difference between them, which gives you only what is not shared between the channels. There is a correlation meter to tell you in advance whether summing will hollow out your file, and you can download the mono version to use elsewhere.
Short loops work best. A two second loop shows its drift within a minute. A three minute file needs a much higher spread before anything audible happens.
Setting the spread
Voices is how many copies you want. Slowest and Fastest set the range of speeds, and everything in between is distributed evenly across the voices. The chips underneath show the exact rate each voice ends up with.
Small numbers do a lot. A spread of +0.05% to +0.6% is already dramatic over a few minutes. Push past 1% and you start hearing it as pitch rather than timing, which is a different and also useful effect.
The preview spreads the voices across stereo so you can hear what is happening before you commit. The render keeps them separate.
Exporting
Choose an output layout:
Objects is the flexible option and the one to start with. It gives you one discrete mono channel per voice, from 2 to 16, with no assumed speaker positions. You place them yourself in your DAW.
Bed layouts (5.1, 7.1, 5.1.2, 5.1.4, 7.1.2, 7.1.4, 9.1.6) assign one voice to each speaker position and write the file with the correct channel order and identification, so it imports as a recognised multichannel file rather than a stack of anonymous channels. The export panel prints the exact channel order before you render.
The LFE never receives a voice unless you ask it to, because it is a band limited effects channel rather than a position in the room. If you do enable it, the signal is low passed at 120Hz as it should be, which means only Pulse or a bass heavy sample will put anything meaningful there.
Everything renders as a 48kHz 24 bit interleaved WAV, faster than real time. The panel shows the file size before you commit, and warns you when a render is large enough that the browser may struggle. Twelve channels at two minutes is about 200MB and comfortable. Sixteen channels at ten minutes is not.
Getting it into an Atmos session
Once the WAV downloads:
- Import the file. In Pro Tools, File then Import then Audio. Choose Split to Mono and you get one mono track per voice, all starting at the same point and already in sync with each other.
- Make them objects. Set each track’s output to an object rather than a bus, then confirm they are showing in the Dolby Atmos window. In Pro Tools Ultimate the renderer is built in, so there is nothing else to launch.
- Place them. Give each track a position with the object panner. This is the part worth spending time on. Voices that are close in rate stay together for a long time, so putting neighbouring rates in opposite corners tends to sound better than putting them side by side.
- Or use it as a bed. If you rendered a bed layout, import it as a single multichannel file onto a matching track instead. The channel order is already correct.
Logic, Nuendo, DaVinci Resolve and Reaper will all take the same file. The only step that differs is how each one assigns a track to an object.
One useful trick
Render longer than you need. Fifteen minutes of twelve voices drifting will contain two or three passages that are far better than anything you would have programmed, because the process finds relationships you would not have thought to try. Bounce it, listen through, and cut out the ninety seconds that work.
Set it running and leave it. That is the point.
