iZotope has introduced Photon, a standalone application and ARA plugin that combines AI-powered stem separation with audio enhancement. Designed for music producers, remixers, podcast editors, and post-production workflows, Photon can separate instruments, isolate speakers, and clean up dialogue within a single application.
Powered by what iZotope describes as 12 custom-trained machine-learning models, Photon processes audio locally on the user’s GPU without uploading files. The launch also marks the first collaboration between iZotope and Boris FX, bringing together the two companies’ research teams.
AI Stem Separation for Music, Drums, and Dialogue
Photon’s Demix models give users several ways to break an existing recording into separate components. For music production, that could mean extracting a vocal for a remix, creating an instrumental, or adjusting elements within a finished mix when the original multitrack session is unavailable.
The Music – 4 Stems Ultra model provides four-stem music separation, while Voice – 2 Stems separates voice from music for quick acapella and instrumental exports. More detailed rhythmic adjustments are handled by Drums – 5 Stems, which separates kick, snare, hi-hat, toms, and cymbals. This gives producers a way to target individual drum elements, such as changing the kick level without applying the same adjustment to the rest of the kit.
Photon also extends stem separation into spoken-word and audiovisual editing:
- Voice Separator: Separates up to six speakers from one audio file into individual stems.
- Cinema – 4 Stems: Separates audio or video into dialogue, music, sound effects, and a fourth stem containing everything else.
For podcast and interview editors, speaker separation offers a way to work on individual voices within a shared recording. Cinema separation provides a similar starting point for editors who need access to dialogue, effects, or music embedded in a combined soundtrack.
Dedicated Models for Dialogue Cleanup
Alongside its stem separation tools, Photon includes Denoise models designed around different speech-recording conditions. Voice Restore addresses several aspects of spoken-word enhancement, including noise, levels, microphone tone, and reverb. Rather than focusing solely on background noise, it is intended to improve the overall presentation of a voice recording.
Photon also includes Indoor and Indoor Noisy models for room noise and reverberation, plus Outdoor, Outdoor Traffic, and Outdoor Wind models for recordings affected by environmental interference. These options let editors choose a model suited to the recording’s specific problems, whether that involves a reverberant interior, passing traffic, or wind around a microphone. Actual results will depend on the source material.
ARA Plugin Integration for DAW Editing
Photon works as a standalone app, but its ARA plugin also brings stem adjustments and background-noise cleanup into compatible DAW workflows. According to iZotope, users can run several models together and switch between them without waiting for each change to process again. That flexibility can help when comparing cleanup approaches or adjusting a stereo recording’s balance during an editing session. The standalone workflow follows a straightforward process: choose a model, process the audio, and export the result.
Local GPU Processing and Fast Previews
A central feature of iZotope Photon is its use of local GPU processing. Its machine-learning models run on the user’s computer, keeping audio files on the system without requiring uploads. Photon also allows processing to begin at any point within a clip. Users can preview a relevant section and, according to iZotope, hear results within seconds. For long interviews, podcasts, or video recordings, this provides a practical way to assess a model before processing the full file. A built-in Model Manager lets you download new and updated models without reinstalling the application.
Audio and Video Export Options
Photon supports exports in WAV, FLAC, and MP3, accommodating both production files and compressed listening copies. For video workflows, its same-as-source video export preserves the original video stream while delivering the processed audio. With music separation, individual drum stems, speaker isolation, and environment-specific dialogue cleanup, Photon brings several common restoration and editing tasks into one package. Its combination of standalone operation, ARA integration, and local processing positions it as a versatile option for creators working across music and spoken-word production.
Photon from iZotope is priced at $379, with cyber deals running through December 7th for $259. You can find it on iZotope’s website here.
