iZotope PHOTON: review of the AI plugin for voice, video cleanup and stem separation
Review of the AI plugin iZotope PHOTON: voice cleanup from noise and reverb, splitting tracks into stems and drums right in the DAW. Is it worth your attention?
Publication date: November 4, 2026
iZotope PHOTON is a recently released plugin that combines several audio processing functions. According to the author's description, it is four tools in one. The plugin works over the ARA protocol (ARA 2 version). It separates material into stems, removes room noise, background noise, and reverb from vocals, separates stems for film, and cleans up recordings for interviews.
Interface and Model Store
The plugin opens in ARA mode. The model store button shows:
- models that are not installed, with their size and purpose;
- models that are already installed, with a description of each.
If the interface is in English, it can be switched to Russian. The store then displays descriptions in Russian.
Example model: "in-room". It reduces background noise and room hum and provides professional voice quality.
Available Models
The author highlights two models:
- Voice Restore: the first one in the list. The author applied it to their own voice.
- Music 4 STS Ultra: the second model. According to the author, the result is impressive.
All models are version 1. The author states that the developer will update them.
Voice Processing
The demonstration was carried out on a recording of the phrase: «Your phone is turned off, and I'm on edge. You think you're not alone, but you're on edge». The author compared the original recording with the processed one by switching the effect bypass.
The controls consist of the following elements:
- Dry/Wet: the first control. It determines the share of the processed signal in the final mix. The scale runs from 0 to 100%. A value of 100% means full processing.
- Room: room reverb control.
- Noise: noise control.
- Tone: tone control. According to the author, it changes the timbre of the signal.
- Level: volume indicator. It shows how much the processing has reduced the signal level. The volume can be returned to the original level.
For a recording with strong reverb and noise, the author sets the Room and Noise controls to 100%, then adjusts the result. The waveform in the lower part of the window shows changes in the signal. In the author's opinion, this is convenient for users who work with waveforms.
Order of actions:
- Perform the processing in the plugin.
- Copy the clip together with the audio.
- Bounce the clip using Command+B.
- Get the processed voice on the track below.
Removing Background Noise from a Video
The author processed a video recording of a dog running along a forest road. In the background, a person on a horizontal bar, a passing car, and forest noise can be heard. The author chose a model that removes noise, turned it on, and got the result. The changes are visible in the waveform image.
Stem Separation from a Movie
The author processed a clip from the film "Spider-Man" with the line: "Jean, your powers are active again". The Cinema 4 SMS model was chosen for processing. It separates the recording into tracks, such as vocals and music. The author notes that the separation was imperfect. According to him, this is due to the low quality of the selected clip.
The tracks can be soloed or deleted.
Music Stem Separation
To separate music, select the Music model, then the four stems mode (four stems) with Ultra quality. Processing runs in real time and online, right in the arrangement window. No file dragging is required.
According to the author's description:
- the bass is isolated fairly cleanly; the bass track can be duplicated, soloed, and bonsai'd;
- the algorithm preserves some of the transients;
- the vocals are not damaged during separation.
Drum Separation
DRUMS mode isolates the drum track from the entire composition. The plugin then separates it into individual elements:
- kick drum (kick);
- snare drum (snare);
- toms;
- hi-hat;
- high-frequency percussion instruments.
The author extracted the hi-hat and the snare. In his assessment, the result is good.
Example of use. The author loaded a reference track into Studio One, ran the processing with an assigned hotkey, and got a separate snare track. He reduced it to a bonsai and used it to study the sound's structure. The author notes that this process takes little time.
Drawbacks
- Bugs. The author notes the presence of bugs. He explains them by the fact that this is the first version of the plugin.
- Freezes. The interface periodically freezes. For example, after opening a new fragment, the window continues to show drums. The user has to manually find the needed section. The author believes the developer may fix this problem in future versions.
Author's Conclusion
The author plans to use PHOTON in every project. Use cases include:
- creating beats;
- analyzing the arrangement of reference tracks;
- extracting vocals;
- isolating individual elements of the drum part.
Introduction and Plugin Capabilities
The author presents PHOTON as a universal AI plugin that works through ARA and combines four functions: stem separation, removal of reverb and noise, room cleaning, and voice preparation for interviews.
- The plugin was released recently and works through ARA (ARA 2).
- Functions: premium-quality stem separation, removal of room tone, noise, and reverb from vocals, stem separation for films, and cleaning up interview recordings.
- The author notes impressive capabilities and a convenient design.
Model Store and Model Selection
The plugin includes a model store that describes the weight and purpose of each model; the interface language can be switched to Russian. The author demonstrates the Voice Restore model for voice.
- The store shows available models for installation with their weight and description.
- Installed models are displayed separately.
- The interface can be switched to Russian.
- The "in-room" model removes background noise and room hum, giving the voice a professional sound.
- Voice Restore is the model the author considers the best; it is applied to his voice.
- Music 4 STS Ultra is the second model; all models are from the first version and will be updated.
Voice Processing Setup
Demonstration of voice processing: the dry/wet control sets the share of processing, and the waveform indicator shows the volume level; at 100% processing is at its maximum.
- The first control, dry/wet, determines the share of the processed signal.
- The indicator shows the volume level and its change after processing; the original level can be restored.
- A value from 0 to 100 represents the degree of processing as a percentage.
- With strong reverb and noise, the controls are set to 100 and then adjusted.
- The processed voice can be copied together with the audio (Command B), producing the result under the original track.
Noise Removal in Video
The author applies noise removal to a video featuring a dog, a person training on a pull-up bar, and street sounds, having chosen a model that removes noise; the result is output quickly.
- Video with a dog, a person on a pull-up bar, a car, a road, and a forest.
- The chosen model removes noise.
- Processing is performed quickly; the result is shown in the image.
Stem Separation for Film
Separation of a track from the film "Spider-Man" into voice and music using the Cinema 4 STS model; the author rates the separation quality as weak due to an unsuccessful fragment.
- A fragment from the film "Spider-Man" was used.
- The Cinema 4 STS model was chosen for separating voice and music.
- The separation turned out not very good; the fragment was unsuccessful.
- The tracks can be soloed or deleted.
Separation into Four Stems
Real-time separation of a music track into four stems (ultra) in the arrangement window; bass and vocals are isolated cleanly, transients are preserved.
- The "four stems ultra" mode is selected, and processing runs online in real time.
- The bass is isolated fairly cleanly, and it can be duplicated and output as a separate track.
- The model preserves some transients and does not damage the vocals.
- No file dragging is required: everything happens in the arrangement window.
Drum Separation
A dedicated function isolates the drums from the whole song; you can extract the kick, snare, toms, and hi-hat. Example with the snare from a reference track in Studio One.
- The DRUMS button isolates the drum track from the whole song.
- You can separately extract the kick, snare, toms, hi-hats, and other high-frequency percussion.
- Example: the snare is isolated from a track in Studio One using a hotkey and saved.
Plugin Shortcomings
The author points out the shortcomings of the first version: bugs and periodic interface freezes, due to which open sections are displayed incorrectly. The author considers them forgivable for a first version.
- Periodic freezes: when opening the needed section, a different section is shown.
- It is inconvenient to search for the needed windows manually.
- The author hopes the problems will be fixed in future versions.
- Bugs are acceptable for a first version.
Conclusions and the Author's Verdict
The author states that they will use the plugin in every project: when creating beats, analyzing reference tracks, and extracting vocals. They invite readers to share their opinions in the comments.
- The author will use the plugin in every project.
- Uses: beat creation, arrangement analysis of reference tracks, vocal extraction.