MIMIC PARTY · DEVELOPMENT NOTE
Why the score can stay in your browser
Mimic Party was designed around a short, local listen–record–compare loop. This note explains the signals the game uses, the limits of the score and the data boundary around a take.
The input is a short take
A scored round begins only after the microphone check succeeds. The browser records a bounded window, keeps the raw samples in the active tab and computes the comparison there. When the round changes or you leave, the active recording is released. Solo play and public rooms do not upload the voice take, and a numeric result can be shared without sending the waveform.
Four useful comparisons
The score combines relative melody, rhythm, attack timing and broad sound shape. A player can hum in a different register and still follow the contour. A louder recording is not automatically better, and a recognizable voice is not the goal. The result card presents the number as a casual game signal, while the replay controls help a player decide what to try next.
Quality checks protect the feedback
Very quiet, very short or clipped input can make a comparison misleading. The game reports a reason-specific retry prompt instead of presenting a confident number for a broken take. A player can retry and keep the run's progress. These checks also make it easier to distinguish a microphone problem from an imitation that simply needs practice.
What telemetry is separate
The site can record bounded gameplay events and numeric results to understand which screens work, but it does not add raw samples or derived voice features to those events. Optional public leaderboards accept the nickname and numbers a player submits. Optional private-room sharing is a separate, explicit action with a short-lived WAV and an expiry rule described in Privacy.
How to read the waveform
The yellow track is the reference and the green track is the take. Their bars show loudness over time, so a player can see whether an attack started late or a phrase stopped early. They are not a pitch chart and cannot identify a person. Listening to the reference and the take remains the best way to choose a new approach.
Why local processing is a product choice
Keeping the comparison in the browser reduces the amount of sensitive material that needs to cross the network and lets a player retry without waiting for a server response. It also makes the limit visible: a local score can be affected by a microphone, room and browser. The game describes it as casual feedback so a number is not mistaken for a voice profile.
A result has a lifecycle
The raw samples exist for the active round and replay controls, then are released when the round changes or the player leaves. Numeric progress and bounded gameplay events have a separate retention story in Privacy. A public leaderboard entry is another deliberate action. Separating these lifecycles keeps a score, a session and a recording from becoming one opaque record.
When a future model changes
If the scoring recipe changes, the public copy should say which signals remain comparable and which results should be treated as a new version. A better algorithm does not make old casual scores verified competition. The useful promise is that the player can hear the reference, understand the feedback and choose another take without surrendering a raw recording.