SMART DEREVERB 1.3.2
Created by Andrea Di Tella
https://www.andreaditella.com

INSTRUMENT & SPEECH DE-REVERBERATION

Free to use with no limitations. Optional donations support development:
https://www.paypal.com/paypalme/andreaditella

INSTALLATION - WINDOWS 64-BIT

Close your DAW or audio/video editor. Extract this ZIP and copy the entire
SmartDeReverb.vst3 FOLDER into:
    C:\Program Files\Common Files\VST3
Administrator permission may be required. Keep the VST3 folder and all of its
contents together; do not copy only the internal file. Restart your host and
rescan its VST3 plugins if needed. Find "Smart DeReverb" in the effects list.
Windows 10 or newer and a compatible 64-bit VST3 host are required.
The C++ runtime is included; no separate Visual C++ runtime installer is needed.

WHAT IT IS FOR

Smart DeReverb reduces unwanted room tails in recorded individual instruments
and speech. It is intended for sources such as harmonica, guitar, solo vocals,
spoken voice, interviews and other individual recordings with excessive room
sound. It is not designed to separate instruments in a complete music mix.

All analysis and processing run locally using conventional digital signal
processing. No AI, neural network, training model, account or cloud service is
used. The plugin estimates late reverberant energy separately across frequency
regions, then applies gradual attenuation while trying to retain direct sound.

The result depends on the recording. Early reflections, room coloration and
reverb overlapping the original sound may remain. Strong processing can affect
natural instrument decay, sustained notes or speech timbre. Aim for a cleaner,
more natural recording rather than complete removal at any cost.

QUICK START

1. Insert the plugin on an individual instrument or speech track.
2. Start with REDUCTION around 25-45% and PRESERVE around 75-90%.
3. Set DECAY to an approximate room tail length, or use LEARN as described below.
4. Raise REDUCTION gradually until the room sound is reduced enough.
5. Compare with BYPASS and hold LISTEN REMOVED to check what is being taken away.
6. If notes or words lose their natural character, lower REDUCTION or SENSITIVITY,
   raise PRESERVE, and refine DECAY.
7. Set OUTPUT as needed, release LISTEN REMOVED and check BYPASS before exporting.

FIVE MAIN CONTROLS

REDUCTION: 0-100%. Amount of reverb attenuation. 0% preserves the aligned input
  signal, apart from OUTPUT gain. High values can sound less natural.
DECAY: 0.10-3.00 seconds. Approximate time for the observed room tail to decay
  by 60 dB. Match the recording; a longer value is not automatically better.
SENSITIVITY: 0-100%. How much preceding signal energy contributes to the
  reverberation estimate. Higher values generally make processing stronger.
PRESERVE: 0-100%. Protection of attacks and steady tonal content. Higher values
  give a more cautious treatment and may leave more room sound. Particularly
  useful for sustained instruments such as harmonica and for natural speech.
OUTPUT: -12 to +12 dB final level. Red L/R LEDs indicate output clipping.

LEARN - CONVENTIONAL DECAY ESTIMATION

Press LEARN while playing a passage that includes clear room tails and pauses.
Keep playback running for about eight seconds. The plugin examines decaying
energy and, if it finds a consistent tail, updates DECAY. The other controls
are kept. The existing treatment continues during learning.

For instruments, use a passage where the player has stopped and the room tail
is exposed. Learn estimates the observed decay; a naturally decaying note can
be confused with room sound. Listen and refine the estimate manually.

If the passage is silent or contains no clear exposed tail, the previous Decay
value is kept and the status explains why. You can always set Decay manually.
Press CANCEL to abandon learning without replacing your settings.
Collection follows audio received from the host. Some hosts send silence when
playback is stopped, so keep the desired section playing until collection ends.

LISTEN REMOVED

Hold LISTEN REMOVED to hear the aligned input minus the treated signal, before
OUTPUT gain. This can include both estimated room sound and any original
instrument/speech content affected by the treatment; it is not an isolated,
perfectly separated reverb stem. OUTPUT also controls the monitoring level.
Release the button to return to normal audio. The audition is temporary and
is not saved with the project. Closing the window stops it. Hosts that signal
an offline render also suppress the audition automatically.

BYPASS AND TIMING

BYPASS compares with the aligned original signal, including its original level.
OUTPUT gain and Removed audition are bypassed as well. Switching is smoothed.
The spectral process introduces a fixed latency at each sample rate, typically
about 43-46 ms at common music sample rates. This is reported to the host.
Enable your host's plugin delay compensation. The same timing is maintained
when Reduction is zero or the plugin is bypassed. It is primarily a tool for
working on recorded audio, rather than for zero-latency live monitoring.

DISPLAY AND OTHER CONTROLS

The spectrum shows INPUT in grey, OUTPUT in amber and REMOVED in muted red.
REMOVED is displayed before OUTPUT gain. It visualizes what the treatment
attenuates, rather than identifying a guaranteed reverb-only signal.
The curves remain a view of the normal treatment while comparing with Bypass
or listening to Removed; the monitoring buttons do not swap the plotted curves.
The display uses a logarithmic frequency axis. There are clipping LEDs and
no VU meters. The window can be resized.

Double-click a knob to restore its default. Click its value to type it.
ABOUT shows the version, author and clickable website. DONATE opens PayPal.

MONO, STEREO AND SAVING

Mono input/mono output and stereo input/stereo output are supported.
Stereo uses the same attenuation pattern for both channels, helping preserve
their balance and avoiding unwanted crossfeed. Host 32-bit and 64-bit floating
point audio streams are supported.

The five controls and Bypass can be automated and are saved with the DAW project.
The learned Decay is an ordinary saved parameter. Learning progress, meter
history and temporary Removed monitoring are not saved.

Website: https://www.andreaditella.com
Donate: https://www.paypal.com/paypalme/andreaditella
