wassermanproductions/stem-studio

Stem Studio is a desktop app for filmmakers and video editors that splits a mixed down soundtrack back into its separate parts. Many delivered videos and reference cuts arrive as a married mix, meaning dialogue, music, and sound effects are all baked onto one audio track, which makes it hard to rebalance the sound, swap out the score, or duck the music under someone speaking. Stem Studio takes that single track and uses local AI stem separation to pull it apart into three clean stems: dialogue, music, and sound effects. You drop in a video or audio file, and the app extracts and normalizes the audio, runs the separation engine while showing live progress, and writes out broadcast ready WAV files at 48 kHz and 24 bit. Alongside the three stems, it also writes a conformed copy of the full original mix in the same format so all four files line up sample for sample, and it states the three stems always sum back to the original mix exactly, with nothing lost or duplicated. If the input was a video, it can also export a version of the original picture with the three stems attached as separate labeled audio tracks, ready to drop into any video editor. Everything runs locally on your own machine, with no upload and no account required. It is available for macOS, Windows, and Linux, installable with a one line Terminal command on Apple Silicon Macs or as a downloadable installer on Windows and Linux. It is one of three connected filmmaking tools from the same author, alongside Motion Previs Studio and Blockout, meant to be used together across a production pipeline. It is released under the Apache 2.0 license.