Published: 11/3/2026
Source: iZotope
iZotope has launched Photon, a standalone application and ARA plugin that runs 12 machine-learning models locally for stem separation, voice isolation, dialogue cleanup, and noise reduction.
Photon can split a song into vocals, drums, bass, and other elements; divide drum audio into five parts; separate as many as six speakers; and isolate dialogue, music, and effects from film audio.
Processing runs on the user’s GPU rather than uploading audio to a cloud service. That makes Photon especially relevant for unreleased masters, client sessions, and recordings covered by confidentiality agreements.
Why it matters: For producers, local AI removes upload queues and usage credits while keeping sensitive audio on the workstation. The practical tradeoff is hardware: compatible Apple silicon or a Windows GPU with at least 4 GB of VRAM is required.