k2k audio A single spectrogram can tell many different stories depending on how you color it. Color modes re-paint the same data to reveal amplitude, pitch, transients, phase damage, or harmonic structure — like switching between X-ray, infrared, and visible light on the same scene.
Deep Analysis Modes
These pre-compute across the full buffer. Slower to calculate, but reveal structure that frame-by-frame modes cannot see.
Transients
What it reveals Attack points and onset energy — every drum hit, pluck, consonant, or bow change lights up against a dark background.
When you’d switch to it You have a dense mix or a layered recording and need to see exactly where the rhythmic energy lives. Useful before applying smoothing or time-domain edits, so you know what you’re about to affect.
Quick example
- Load a drum loop or full mix.
- Switch the color mode to Transients.
- Bright vertical stripes appear at each attack; sustained tones stay dark.
- Raise Sensitivity to catch softer ghost notes, or lower it to isolate only the hardest hits.
- Increase Duration to widen the bright region around each onset — helpful for seeing the full transient tail.
Parameters
| Parameter | What it controls | Range | Default |
|---|---|---|---|
| Sensitivity | How strong an onset must be to register | 0.00 — 1.00 | 0.50 |
| Duration | How many milliseconds the highlight extends after each onset | 5 — 100 ms | 20 ms |
| Min Interval | Minimum gap between two detected onsets | 10 — 500 ms | 50 ms |
Color mapping Dark where the signal is sustained or silent. Brightness ramps from deep teal through cyan to yellow at each detected onset, fading over the duration window. A very hot transient pushes into red.
Harmonics
What it reveals Sustained harmonic peaks averaged across the entire buffer — the stable pitched bones of the sound.
When you’d switch to it You want to see where the strong resonant frequencies sit before sculpting, EQ-ing, or extracting harmonic content. Particularly revealing on instruments with clear partials: piano, strings, voice, brass.
Quick example
- Load a sustained note or chord.
- Switch to Harmonics.
- Horizontal bands glow at each harmonic partial; noise and transients stay dark.
- Widen Peak Width to blur the bands together (useful for dense harmonic clusters) or narrow it for surgical precision.
- Lower Threshold to reveal quieter upper partials.
Parameters
| Parameter | What it controls | Range | Default |
|---|---|---|---|
| Peak Width | How wide each detected peak glows, in Hz | 10 — 500 Hz | 100 Hz |
| Threshold | Minimum average level a peak must reach to be shown | -80 — -20 dB | -60 dB |
Color mapping Dark where no harmonic peak is detected. Brightness follows the continuous ramp — teal near the edge of a peak, through cyan, to bright yellow at the peak center. Bins well outside any detected peak stay black.
HPSS
What it reveals The balance between harmonic (sustained, pitched) and percussive (transient, noisy) energy in every time-frequency bin, shown as two opposing colors in a single view.
When you’d switch to it You need a quick read on how much of the signal is rhythmic versus tonal — before running an extractor, or to verify that an HPSS-based separation is doing what you expect.
Quick example
- Load a full mix or a loop with both drums and melody.
- Switch to HPSS.
- Tonal elements (vocals, pads, bass notes) glow in one hue; percussive elements (snare, hi-hat, pick noise) glow in the opposite hue.
- Increase Harm. Kernel to broaden what counts as harmonic — long sustained tones become more dominant.
- Increase Perc. Kernel to broaden what counts as percussive — short bursts claim more territory.
Parameters
| Parameter | What it controls | Range | Default |
|---|---|---|---|
| Harm. Kernel | Median filter length along time (must be odd) — larger values favor longer sustained sounds | 3 — 63 | 31 |
| Perc. Kernel | Median filter length along frequency (must be odd) — larger values favor broader transient bursts | 3 — 63 | 31 |
| Mask Power | Sharpness of the harmonic/percussive separation — higher values push bins toward one side or the other | 1.0 — 4.0 | 2.0 |
Color mapping Bins dominated by harmonic energy appear in cyan; bins dominated by percussive energy appear in warm salmon-orange. Where the two are roughly equal, the color passes through a violet midpoint. Brightness tracks overall intensity — quiet bins stay dark regardless of their harmonic/percussive ratio.
Formants
What it reveals Vocal tract resonances (F1 through F4) and other broad spectral envelope peaks — the shapes that give a sound its vowel-like character.
When you’d switch to it You are working with voice, woodwinds, or any source where the spectral envelope matters more than individual partials. Useful for spotting vowel transitions, checking formant consistency across takes, or previewing where a formant-shift effect will act.
Quick example
- Load a vocal phrase.
- Switch to Formants.
- Broad horizontal bands glow at the formant frequencies, shifting frame by frame as vowels change.
- Raise Cepstral Order for a more detailed envelope (more peaks resolved); lower it for a smoother, broader view.
- Widen Formant Width to spread each detected resonance across a broader frequency region.
Parameters
| Parameter | What it controls | Range | Default |
|---|---|---|---|
| Cepstral Order | Number of cepstral coefficients used to estimate the spectral envelope — higher values capture finer detail | 4 — 80 | 40 |
| Formant Width | How wide each detected formant peak glows, in Hz | 50 — 500 Hz | 200 Hz |
Color mapping Dark outside detected formant regions. Brightness follows the continuous ramp — teal at the edges of each resonance, through cyan to yellow at the formant center. Only peaks within the vocal range (roughly 80 Hz to 6 kHz) are highlighted; everything outside stays dark.