Speech Frequency Spectrogram

Many times a second, the tool runs a Fast Fourier Transform (FFT) on the latest slice of your microphone signal, giving the energy at each frequency. It paints that as a vertical column of coloured pixels — low frequencies at the bottom, high at the top on a logarithmic frequency axis — and scrolls the image left so a continuous real-time spectrogram builds up over time. Brighter pixels mean more energy. In speech you’ll see distinct horizontal bands (the formants that shape vowels — particularly F1 around 300–800 Hz and F2 around 800–2500 Hz), a striped low region where the voice’s harmonic series sits, and brief vertical smears for plosives (bursts at “p”, “t”, “k”) and the high-frequency fuzz of fricatives like “s” and “sh”. Everything runs live on your device; nothing is recorded.

A spectrogram shows sound; it doesn’t read it. This tool does not transcribe words or tag phonemes — that’s speech recognition, a different problem requiring machine-learning models. Use it as a voice spectrogram to study how vowels and consonants look in the time-frequency domain, to explore phonetics by comparing different vowel sounds, to see how nasality or breathiness affect the harmonic structure, or to watch how your pitch shifts while singing. If you also want to inspect which harmonic overtones are present in your voice, the harmonic overtone detector lists them numerically alongside this visual view. The picture also depends on your microphone’s response and the room’s acoustics — a boomy room will smear the low end, and a bright mic will intensify the upper bands.