Sonible smart:chain: AI mixing setup and workflow
Learn how smart:chain learns tracks, balances groups and adjusts mix depth, with local processing and up to 24 manual EQ bands.
Player not loading? Watch on YouTube
Sonible's Alex demonstrates smart:chain as a channel-strip plug-in whose instances share information across a DAW session. Its seven modules run in a fixed order: gate, EQ, de-harsher/de-esser, input rider, compressor, saturation and clipper. He explains that the order lets the compressor respond to EQ decisions.
The setup starts with an instrument profile for each track. Alex selects a passage where as many sources as possible are playing, then starts the learning phase. He warns against learning from microphone bleed rather than the intended instrument. Optional auto-leveling provides an initial balance; group trim and individual controls allow further adjustment. Front/mid/back placement changes EQ and compression, while the EQ also supports up to 24 manual bands.
For viewers seeking local AI audio tools, Alex says the neural networks run entirely on the computer and audio data does not leave it. The presentation lists VST3, Audio Unit and AAX support on Mac and Windows.
The workflow assumes separate instrument tracks rather than buses or several instruments sharing one track. Alex gives a limit of 10 members per group and says overall capacity depends on CPU resources. He describes music multitracks as the training focus, although speech processing is possible. The demonstrated mute/solo and revised group controls belong to a prerelease update during the recording.