iZotope and Boris FX Launch GPU-Accelerated Audio App Photon
iZotope released Photon, a standalone desktop application and Audio Random Access plug-in that utilizes 12 custom-trained machine learning models for AI-powered stem separation and audio enhancement, Mixonline reported. Developed as the first collaboration between iZotope and Boris FX, the software runs locally on computer graphics processing units to process audio files without cloud uploads.
Photon provides local GPU processing designed for fast, private rendering on hardware configurations meeting minimum system requirements. Red Shark News reported that the software requires macOS 14 or later running on Apple silicon, with Intel Macs unsupported, or a 64-bit Windows 10 or later operating system paired with a GPU featuring at least four gigabytes of VRAM. The application recommends 16 gigabytes of system memory, setting eight gigabytes as the bare minimum.
Isolate audio elements using specialized machine learning models
The application divides its 12 machine learning models into specific categories for isolating elements and cleaning up audio recordings. Music production features include Demix Music 4 Stems and a new Music 4 Stems Ultra model. A dedicated Drums 5 Stems model separates individual elements including kicks, snares, hi-hats, toms, and cymbals. For spoken audio, the Voice Separator handles up to six distinct speakers from a single file, while a Voice 2 Stems model splits voice from music. Cinema 4 Stems divides video or audio files into dialogue, music, sound effects, and ambient noise.
For background cleanup, the Denoise suite includes Voice Restore for speech, microphone tone, and reverb management. Indoor models target room noise and reverberation, while outdoor configurations focus on traffic, urban environments, and heavy wind.
Run models inside workstations with flexible export formats
As an ARA plug-in, Photon allows engineers to run several models simultaneously and switch between them inside digital audio workstations without waiting for renders. Users can start previews at any point in an audio clip. A built-in Model Manager lets operators download new and updated models without reinstalling the software. Export options include WAV, FLAC, MP3, and video formats that preserve the original video stream while rewriting the audio track. Supported input file formats span .wav, .mp3, .flac, .aiff, .aif, .aac, .m4a, .mp4, .mkv, .mov, .mxf, and .mpg.
Photon is available now through izotope.com/products/photon, featuring early cyber pricing deals running through December 7, 2026. Standard manufacturer’s suggested retail pricing is $349, with a promotional cyber price of $259. Crossgrade tiers are priced by existing software ownership: $269 ($179 cyber price) for users crossing over from any paid iZotope product; $199 ($129 cyber price) from RX Elements or bundles containing it; $149 ($74 cyber price) from RX Standard or associated bundles; and $99 ($49 cyber price) from RX Advanced. Subscription options cost $19 per month ($12 for the first month, then $19) or $119 per year ($89 for the first year, then $119).
Frequently Asked Questions About Photon
What types of video files can Photon open and process?
Photon opens MP4, MKV, MOV, MXF, and MPG files, reads their existing audio tracks, and writes the processed audio back into the file while keeping the original video stream intact.
How does the Model Manager update the software?
The Model Manager downloads new and updated machine learning models directly within the application, eliminating the need to perform full software reinstalls.
Does Photon support older Intel-based Mac computers?
Intel Macs are not supported by the software, which requires macOS 14 or later running exclusively on Apple silicon.