Spectral repair is a tool for interpolating selected areas on a time-frequency spectrogram. It is able to provide higher quality than the Declicker tool for long corrupted segments of audio (above 10 ms). Spectral repair can be used to remove (or attenuate) certain unwanted sounds from recordings, such as squeaked chairs, coughs, dropped objects, mobile phone calls, etc.

Attenuate: this mode can be used when the corrupted interval contains sufficient
useful information. This method reduces spectrogram magnitudes in the selected
area to match magnitudes from the surrounding area.
Replace: this mode
completely replaces the selected content with a content interpolated from the
surrounding data. The number of frequency bands used for interpolation is
selectable.
Pattern: this mode finds the most similar portion of the
surrounding audio and uses this to replace the corrupted audio.
Partials + Noise: this mode allows for higher-quality interpolation by explicit location of
signal harmonics from 2 sides of the corrupted interval and linking them
together by synthesis. This method is able to correctly interpolate cases of
pitch modulation, including vibrato. The rest of non-harmonic material
("residual") is interpolated using Replace method.
Controls:
Number of bands - selects the number of frequency bands used for interpolation.
A higher number of bands can provide better frequency resolution, but also requires wider surrounding area to be analyzed for interpolation.
Multi-resolution -enables multi-resolution mode when better frequency resolution is used for interpolation of low-frequency content and better time resolution is used for interpolation of high-frequency content.
Strength - parameter adjusts strength of attenuation in Attenuate mode.
Surrounding region length - defines how much of the surrounding content will be used for interpolation
Before/after weighting - gives more weight to the surounding audio before or after the selection
Harmonics sensitivity - adjusts amount of detected and linked harmonics in Partials + Noise mode.
Lower values will detect fewer harmonics, while higher values will detect more harmonics and can introduce some unnatural pitch modulations in the interpolated result.