CEDAR currently offers four voice-focused plugins that can work in real time, but their controls and intended sources point to different engineering jobs. Choosing by brand name alone wastes soundcheck time. Start with the source, the failure you are fixing, and the control you need.
StageVox and ScreenVox use CEDAR Quantum near-zero-latency noise suppression, while Voxis and VoicEX 2 use CEDAR neural network processing. CEDAR specifies less than 10 milliseconds of latency for both neural products. All four tackle background noise or excessive reverb, yet they are not interchangeable.
The launch of CEDAR’s live neural voice isolator added a band-based option for speech and singing. That matters when the unwanted sound changes across the frequency range instead of sitting behind one source type. It does not make the older tools obsolete.
ScreenVox takes the same broad low-latency idea toward speech. Its stated jobs include broadcasting, podcasting, theater, sports studios, and newsrooms, while the Bias control replaces StageVox’s Ambience control. A presenter who needs clear consonants and stable lip sync is closer to ScreenVox’s design center than a singer working beside a loud drum kit.
Voxis becomes more attractive when the source may be speech or singing and the noise has a complicated tonal shape. Its low, mid, and high controls let you distribute isolation by frequency, while Link moves all three together and preserves their balance. The separate Voice control can then raise the isolated voice without pretending every noise problem needs the same treatment.
VoicEX 2 also recognizes speech and vocals, but its interface separates the background and voice into two level decisions. You can reduce or emphasize either component, which suits restoration, broadcast, podcast production, and other work where the foreground-background balance matters. CEDAR now specifies less than 10 milliseconds of latency, so older comparisons calling VoicEX 2 unsuitable for live use are outdated.
Voxis exposes a different decision in its interface. It can treat air-conditioning rumble harder in the low band and audience wash higher up, without forcing equal processing through the midrange. VoicEX 2 offers less spectral steering but more direct control over the relative loudness of voice and background.
That distinction changes a practical plugin comparison. Use the tool whose controls describe the correction you actually need, then compare processed and bypassed signals at matched loudness. Louder output can masquerade as clearer output, especially when a Voice control or foreground level has been raised.
Strong processing also exposes different failure modes. Listen for softened consonants in speech, unstable breath detail, flattened room character, and tonal changes when singing crosses registers. A setting that cleans a quiet verse can react differently when crowd noise and vocal level rise together.
A mixed production does not require one processor on every voice. ScreenVox can handle spoken presenters while StageVox handles singers, leaving Voxis for channels where variable spill defeats the simpler source-specific approach. VoicEX 2 remains useful on a record or stream feed whose voice-to-background balance may need editorial adjustment.
That difference matters most when performers hear the processed path. ScreenVox or StageVox may suit dense per-channel deployment, while the neural products justify their extra latency when isolation quality or control structure solves the harder problem. Measure the complete route under show load rather than treating one plugin specification as the system total.
CEDAR offers 14-day trials for all four products, making a controlled comparison possible before purchase. Feed each plugin the same representative microphone recordings, match output loudness, and judge preserved consonants, breath detail, ambience, and recovery after sudden noise. A repeated phrase after a crowd swell reveals behavior that a clean solo clip cannot.
StageVox and ScreenVox use CEDAR Quantum near-zero-latency noise suppression, while Voxis and VoicEX 2 use CEDAR neural network processing. CEDAR specifies less than 10 milliseconds of latency for both neural products. All four tackle background noise or excessive reverb, yet they are not interchangeable.
The launch of CEDAR’s live neural voice isolator added a band-based option for speech and singing. That matters when the unwanted sound changes across the frequency range instead of sitting behind one source type. It does not make the older tools obsolete.
The source type narrows the field fast
StageVox is the direct fit when the wanted source is a singer and monitoring delay must stay extremely low. CEDAR positions it for live performance, tracking, worship, recording, and mixing, with Attenuation, Focus, and Ambience controls. That last control makes near-zero vocal cleanup useful when you want less spill without erasing every trace of the room.ScreenVox takes the same broad low-latency idea toward speech. Its stated jobs include broadcasting, podcasting, theater, sports studios, and newsrooms, while the Bias control replaces StageVox’s Ambience control. A presenter who needs clear consonants and stable lip sync is closer to ScreenVox’s design center than a singer working beside a loud drum kit.
Voxis becomes more attractive when the source may be speech or singing and the noise has a complicated tonal shape. Its low, mid, and high controls let you distribute isolation by frequency, while Link moves all three together and preserves their balance. The separate Voice control can then raise the isolated voice without pretending every noise problem needs the same treatment.
VoicEX 2 also recognizes speech and vocals, but its interface separates the background and voice into two level decisions. You can reduce or emphasize either component, which suits restoration, broadcast, podcast production, and other work where the foreground-background balance matters. CEDAR now specifies less than 10 milliseconds of latency, so older comparisons calling VoicEX 2 unsuitable for live use are outdated.
The controls reveal four different workflows
StageVox asks how much attenuation you need, what noise should receive attention, and how much natural ambience should survive. ScreenVox substitutes Bias so you can tune the retained voice instead. Those workflows are deliberately source-specific, even though both products can operate outside their headline use cases.Voxis exposes a different decision in its interface. It can treat air-conditioning rumble harder in the low band and audience wash higher up, without forcing equal processing through the midrange. VoicEX 2 offers less spectral steering but more direct control over the relative loudness of voice and background.
That distinction changes a practical plugin comparison. Use the tool whose controls describe the correction you actually need, then compare processed and bypassed signals at matched loudness. Louder output can masquerade as clearer output, especially when a Voice control or foreground level has been raised.
Strong processing also exposes different failure modes. Listen for softened consonants in speech, unstable breath detail, flattened room character, and tonal changes when singing crosses registers. A setting that cleans a quiet verse can react differently when crowd noise and vocal level rise together.
A mixed production does not require one processor on every voice. ScreenVox can handle spoken presenters while StageVox handles singers, leaving Voxis for channels where variable spill defeats the simpler source-specific approach. VoicEX 2 remains useful on a record or stream feed whose voice-to-background balance may need editorial adjustment.
Latency and instance count decide the final fit
CEDAR describes StageVox and ScreenVox as near-zero-latency processors and says both are efficient enough for many simultaneous instances. Voxis and VoicEX 2 carry a sub-10-millisecond processing claim instead. Your interface buffer, plugin host, converters, console path, and network transport still add delay around that figure.That difference matters most when performers hear the processed path. ScreenVox or StageVox may suit dense per-channel deployment, while the neural products justify their extra latency when isolation quality or control structure solves the harder problem. Measure the complete route under show load rather than treating one plugin specification as the system total.
CEDAR offers 14-day trials for all four products, making a controlled comparison possible before purchase. Feed each plugin the same representative microphone recordings, match output loudness, and judge preserved consonants, breath detail, ambience, and recovery after sudden noise. A repeated phrase after a crowd swell reveals behavior that a clean solo clip cannot.
