iZotope's Photon uses multiple models on your GPU to split stems and clean up noisy audio. It's available now as a standalone app and ARA plugin.
iZotope has launched Photon, a standalone app and ARA plugin for AI-powered stem separation and audio enhancement that uses 12 custom-trained models and runs locally on the GPU, with no uploads. Perhaps more significantly than that, for the long term at least, Photon is the first collaboration between the iZotope and Boris FX research teams.
It can split music into four stems, separate drums into individual elements [Handy for trying to work out what on earth Danny Carey is up to this time. Drumming Ed], isolate vocals, identify and separate speakers, and clean up noisy recordings. As an ARA plugin, it lets users run several models together and switch between them. Previews can start at any point in a clip. Model Manager downloads new and updated models without a reinstall.
Exports are WAV, FLAC, MP3, or video in the same format as the source, with the original video stream kept. Photon opens a lot too, namely .wav, .mp3, .flac, .aiff, .aif, .aac, .m4a, .mp4, .mkv, .mov, .mxf, and .mpg files. For video, it reads the audio track and can write the processed audio back.
Photon is available now at izotope.com/products/photon, with cyber deals through December 7, 2026. Prices are MSRP, with cyber prices in brackets.
Photon requires macOS 14 or later on Apple silicon (Intel Macs are not supported), or Windows 10 or later, 64-bit, with a GPU of at least 4 GB VRAM. 16 GB of RAM is recommended and 8 GB is the minimum.