Upload audio, select and drag boxes directly, and modify frequency properties with instantaneous live sound feedback as you edit.
Import your audio track
Supports WAV, MP3, and OGG formats up to 48kHz sampling rate
Select any box to move or drag-resize it. Hold Shift to select multiple boxes.
Lower depth retains harmonic content; -80dB is maximum erase.
Higher Q targets a narrower notch, preserving sidebands.
Live Parameter Tuning
Click any drawn box on the spectrogram to select it and tune its bandwidth, depth, and notch frequencies in real-time.
Decoding bitstream, allocating offline filter array, and generating localized attenuation channels using the Web Audio offline system.
Surgical audio attenuation applied to the specified regions. Your output file has been synthesized locally in the browser cache.
Unlike traditional vocal removers that rely on simple out-of-phase stereo cancellation (which damages spatial stereo signals and wipes out drum transients panned dead-center), our Manual Spectrogram Eraser uses multi-dimensional Short-Time Fourier Transforms (STFT) to convert raw, time-domain audio wave structures into visually identifiable frequency spectra.
To convert a continuous time-varying audio signal $x(t)$ into its mathematical spectrogram layout, we break down the signal into finite, overlapping windows. Each localized window undergoes a discrete Fourier transform computation:
The Auto-Detect (Beta) feature utilizes a localized offline audio context to apply a bandpass biquad filter focusing exclusively on fundamental vocal formants ($300\text{Hz}$ to $3000\text{Hz}$). It calculates Root Mean Square (RMS) energy densities at $1\text{fps}$ granularity to map potential vocal regions.
🔓 Watch this short ad to unlock your download
30s