Scream anything, it becomes a coin: your recording becomes the coin's sound, played by its pool, and its hash is fixed on-chain.
Scream is the fastest way to launch on Pitch. Hold the mic, scream up to five seconds, check the name and ticker, set the opening buy and launch. Your recording is published with the coin and becomes its sound; everything else in the record is filled in from your voice's fingerprint.
What you hear
A screamed frequency plays exactly what you screamed: the take, trimmed to its voiced part (at most five seconds, mono), looped with a short crossfade. The pool plays it. Price sets the pitch: its position inside its own 24-hour log range maps to −12…+12 semitones (the bottom of the range an octave down, the top an octave up, smoothed), so a pump bends the scream up and a dump drags it down. Volume swells the gain and layers two slightly detuned copies on top. Liquidity sets how long it echoes. A limiter caps the output at every step.
| Pool | Sound |
|---|---|
| Price (position in its 24 h log range, 0…1) | Playback pitch: (p − 0.5) × 24 semitones, clamped to ±12, smoothed |
| Trades / 24 h volume | Gain swell (up to +50 %) and doubled layers (±9 cents) |
| Liquidity | Echo length and feedback |
What is measured
- Duration of the voiced part, peak and average loudness.
- A 16-point loudness envelope (how the scream swells and fades).
- Pitch per 40 ms frame (YIN autocorrelation): the median note, the range between the 10th and 90th percentile, and whether it rises or falls.
- Brightness (spectral centroid), roughness (zero-crossing rate) and the share of frames with a stable pitch.
- Bursts: sudden jumps in loudness, and how regularly they are spaced.
Every value is rounded coarsely (pitch to the semitone, brightness to 50 Hz, loudness to 3 dB, durations to 0.1 s) so the same gesture lands on the same numbers. The rounded set is the fingerprint; its keccak-256 hash is the seed. Same fingerprint, same frequency, for anyone.
How it maps to sound
| Your voice | Becomes |
|---|---|
| Three or more evenly spaced bursts | Cellular automaton |
| Short punchy bursts | Physical-modeling resonator |
| A steady pitched tone | Coupled oscillators |
| A pitched voice sliding over a wide range | Granular synthesis |
| A noisy, bright scream with little stable pitch | Stochastic process |
| Median pitch | Root note |
| Rising / falling / level / noisy | Major pentatonic / minor pentatonic / dorian / phrygian |
| Burst rate and pitch range | Tempo bounds |
| Loudness | Maximum simultaneous voices |
| Time to the loudest point / the fade after it | Attack / release |
| Brightness (dark, warm, bright) | Variant of the class |
The rounded fingerprint is stored in the frequency's record under voice, so anyone can read what seeded it and check that the record follows from it. The same block holds the recording: audioUrl, audioSha256, audioMime and audioDurationMs. The token image is drawn from the same fingerprint.
What is published
Content
Scream only yourself. Before launching you confirm that you recorded the take yourself, that it contains no one else's voice and no copyrighted music, and that you accept it being public forever. Recordings of other people, copyrighted material, threats or harassment are not allowed. Every frequency page has a Report link; reported recordings can be hidden from this site, but the on-chain record and its hash cannot be erased.
Names
The transcript becomes the token name in Title Case and the ticker from its first meaningful words ("wen moon" becomes WENMOON). You can always edit both. Slurs, swearing-heavy tickers and obvious brand or celebrity impersonation are refused. Want to set every field yourself? Use Advanced on the launch page.