Nobody can translate cat. What this page does is measure the meow and infer an intent from acoustic features that published studies found correlate with context. The words are personality — his, and by extension, ours.
What it measures
- Fundamental frequency (pitch) by normalized autocorrelation on 43 ms windows, with parabolic peak refinement. Adult cat meows sit around 400–800 Hz; a male human voice around 85–180 Hz. That gap is what lets "Everyone at once" tell you two apart.
- Contour — whether the pitch rises or falls from the first third of the meow to the last. Lund University's Meowsic project (Schötz et al.) found rising melodies cluster in friendly, attention-seeking contexts and falling ones in stressed or complaining contexts.
- Duration, loudness, harmonicity and jitter. Long and loud reads as demand; short and clean as greeting; rough and low as complaint. Nicastro & Owren (2003) showed humans can classify meows by context above chance from features like these, so the heuristics aren't nothing — but they aren't a dictionary either.
- What it can't hear: a purr sits around 25 Hz, which is barely one cycle inside a 43 ms window, so purrs register as quiet noise and are ignored. He'll have to purr at you in person.
How it meows back
A sawtooth source at a cat-range pitch runs through three sweeping band-pass "formant" filters that shape the m → ee → ow of a meow. Your sentiment sets the pitch glide, length, roughness and whether he purrs or trills. Cats do react to synthesized meows at natural pitch — mostly by looking for the other cat.
Privacy
Audio analysis and meow synthesis run entirely in this tab. Transcription of your voice uses the browser's built-in speech engine (on Chrome that engine sends audio to Google's speech service — no keys, nothing stored by this page). If that engine isn't available you get a text box instead. The conversation lives in this device's local storage only.
