Reduce steady hiss in a stereo recording
Admin ·
Use spectral noise reduction when a steady background hiss runs under speech or music. Learn the shape of the noise from the opening of the selected passage or set a fixed floor. The browser attenuates frequency components and saves 16-bit WAV or FLAC.
Compare a quiet pause and a stereo signal
Open the downloadable 32,000 Hz sample. Set Start to 0.125 and End to 0.875. Keep the opening noise reference, choose −32 dB for the floor and 24 dB for reduction, and select WAV. The profile comes from about the first 0.2 seconds of the selection, which contains seeded broadband noise. The tones enter at source time 0.75 seconds.
Listen to the opening pause and the transition. The export lasts 0.75 seconds, has 24,000 sample frames in each channel, and the tones enter near 0.625 seconds on its timeline. In the interior of the opening pause, each channel's measured RMS noise falls below 55% of its unfiltered value. The main 440 Hz and 630 Hz tones retain more than 90% of their measured amplitude in the checked segment.
Download the noisy stereo WAV sampleFFmpeg FFT audio noise reduction filter
Listen for consonants before increasing the floor
For a voice memo, choose a selection that begins with room noise alone, then includes a quiet word. Try the default −40 dB floor and 12 dB reduction first. If you cannot start with an isolated stretch of noise, select the manual floor. Use the manual floor for an export shorter than 0.35 seconds.
Compare at the same listening level. If a quiet consonant warbles or a high note loses its texture, choose a more negative floor or a smaller reduction. The filter's floor is a control for the spectral estimate. Raising the floor and raising the reduction change different parts of the filter.
Choose a 16-bit working copy for review
WAV stores the processed PCM samples directly. FLAC compresses those same 16-bit samples losslessly. Use whichever the receiving editor accepts and keep the untouched recording for another pass. Your file is read and processed on the device.
A nearby conversation, microphone overload and a short impact call for different edits. Compare those sections before applying one spectral setting to the full recording.
Common questions
Will a selected part have the original sample rate?
Yes. The export keeps the source sample rate and channel count and puts the selected passage at the beginning of the new file.
Why check more than a quiet pause?
A pause reveals residual noise. Speech or music reveals whether a setting also reduced parts you want to keep.