# Product Memorandum: CALM EverClean

## Executive Summary

CALM EverClean is a free vocal processing and feedback-control plugin for live sound, broadcast, theatre, conferences, streaming, houses of worship, small venues, and any situation where a voice needs to remain clean, stable, and usable in real time.

The plugin combines two main processing paths:

- **CALM**, a tonal-control and anti-feedback engine with a Studio 21 ms mode and a Zero Latency mode.
- **EverClean**, a zero-added-latency vocal cleaning engine designed to reduce noise, room tail, bleed, and tonal dirt without forcing the engineer into a complex workflow.

The goal is not to replace proper gain structure, microphone choice, microphone placement, or system tuning. The goal is to provide an additional safety and cleanup layer for difficult vocal channels while keeping the workflow fast, practical, and free for the community.

CALM EverClean is distributed as a free beta. The project is driven by real-world feedback from engineers, musicians, houses of worship, venues, theatre companies, broadcast users, and live-tool users.

## Product Philosophy

CALM EverClean is built around a simple idea: advanced live cleanup tools should not be limited to expensive closed systems. In a real show, an engineer needs clear controls, reliable latency, and predictable behavior. For that reason, the plugin avoids forcing every user into a heavy modular workflow and separates operation into two levels:

- **Basic**, for fast access to CALM, EverClean, DNS, Auto Gain, and the latency selector.
- **Advanced**, for users who want the full channel strip: expander, resonance control, clean EQ, advanced DNS, de-esser, body control, compression, and finisher.

The design principle is that the plugin should help without getting in the way. The main controls must be usable in a live situation without opening a large number of menus.

## Main Modules

### CALM

CALM is the anti-feedback and tonal-stability core. In Studio mode it uses spectral analysis and neural control to detect stable tonal components, resonances, and build-ups that may become feedback. In Zero Latency mode it uses a causal narrow-band path to provide immediate protection without lookahead.

Primary uses:

- Reduce feedback risk.
- Control stable tones and resonances.
- Keep a voice more stable on sensitive systems.
- Work in combination with DNS and Auto Gain.

### EverClean

EverClean is the zero-added-latency vocal cleaning path. It processes the signal causally, sample by sample, while maintaining persistent internal states that adapt to the behavior of the channel.

Primary uses:

- Clean background noise.
- Reduce stage bleed.
- Control room tail and tonal dirt.
- Make a voice feel closer and drier when the source allows it.

EverClean is not intended to behave like a fixed EQ. Its behavior is dynamic and depends on the incoming signal.

### DNS

DNS is the noise-control module. In the Basic workflow it can be used quickly with safe settings. In Advanced mode it exposes bias and attenuation controls.

Primary uses:

- Reduce stationary background noise.
- Help vocal channels recorded or amplified in noisy environments.
- Complement CALM and EverClean when a cleaner signal is needed.

### Auto Gain

Auto Gain compensates for level changes caused by cleaning or dynamic reduction. Its purpose is not mastering; it is meant to prevent the user from confusing "lower volume" with "better cleanup."

### Advanced Channel Strip

Advanced mode keeps additional modules for finishing the vocal channel:

- Expanse: dynamic background and room control.
- Resonance: resonance reduction in Studio mode.
- Clean EQ: focus and clarity shaping.
- Advanced DNS: bias and attenuation.
- Body: body/proximity control.
- De-esser: sibilance control.
- SSL-style compression: channel/bus compression.
- Vocal Finisher: drive, opto, and air.

These modules are not required to use CALM EverClean, but they allow the plugin to operate as a complete vocal channel strip.

## Latency Modes

### Studio 21 ms

Recommended when maximum selectivity is the priority:

- Better spectral analysis.
- Stronger feedback and resonance control.
- Suitable for mixing, broadcast, streaming, theatre, recording, and situations where 21 ms is acceptable.

### Zero Latency

Recommended when latency is critical:

- Zero added latency.
- EverClean runs through the causal path.
- CALM uses causal narrow-band feedback protection.
- DNS can run in Basic zero-latency mode.

This mode is intended for live sound, monitors, spoken voice, and situations where latency is the absolute priority.

## Protection and Integrity

EverClean's internal resources are embedded in encrypted form inside the plugin binary and verified on load. This avoids external model installation, reduces the risk of broken installs, and helps prevent accidental modification or missing resource files.

The user does not need to install separate models.

## Formats

The current beta is delivered as:

- macOS AU
- macOS VST3
- macOS Standalone
- Windows VST3
- Windows Standalone

The current macOS beta may be unsigned or not notarized until the Developer ID signing process is completed. The Windows beta may require the user to accept the system security warning if it has not yet been signed.

## Beta Status

CALM EverClean is an active beta. Practical feedback is especially useful:

- Host/DAW used.
- Operating system.
- Plugin format.
- Buffer size.
- Microphone and source.
- Settings used.
- What improved.
- What got worse.
- Whether the user would trust it in a real show.
- Recall, CPU, loading, crash, or automation bugs.

## Positioning

CALM EverClean will remain free. The goal is to build a cooperative, practical, and usable alternative for vocal cleanup and feedback control in live sound.

It is not presented as magic or as a replacement for proper engineering practice. It is presented as a real tool, open to real testing, and improved through real feedback.