Let your voice take flight. Lift is a self-paced game for exploring the beginning of a voiced sound. A verified soft start unlocks takeoff. Your recorded volume then controls the glider throughout the rise, plateau, and descent. The glider moves forward only while your voice matches the current target window.
01
See the connection. Watch a demo. An above-window start stays on the island; a verified soft start unlocks takeoff. Follow the rest of the route with volume; the example also shows a held flight outside the window.
02
Begin when ready. Calibrate with normal speech. Then begin a vowel softly inside your selected amplitude window. There is no countdown.
03
Choose your next flight. Increase volume through the rise, maintain it across the plateau, and decrease it into the soft window to land. Pauses hold your exact position. Repeat the prompt or breathe whenever needed.
A soft beginning, followed by a voice-controlled route. The default takeoff window is 25–60% of the noise-corrected microphone amplitude measured during normal-voice calibration. Select a preset or set your own window. The glider remains still until the voiced beginning supplies 140 ms inside the window. A start above the upper limit cannot be rescued by turning the same sound down: pause and start again. After a valid takeoff, the target window rises to the selected plateau volume and falls back into your soft window. Voice inside the current window advances the glider. Voice outside it, silence, unvoiced noise, or clipping do not advance it. The top of the window never exceeds the calibrated normal reference. These are adjustable game settings, not clinical thresholds.
Vowels and the built-in words and phrases begin with a vowel sound. Custom prompts are not checked. Lift does not recognize words or verify pronunciation.
How to follow the flight The mint corridor is the volume target. The glider’s height responds to smoothed recorded volume; its forward position changes only when your voice is in range. A small ring shows the current target position when the guide is on. At the start and end of the plateau, and at landing, a 120 ms in-window check must finish before the next stage unlocks.
Matched voice time is counted only while your voice is in the window. Turn checks and adjustments add time. Pauses do not make the glider fall or progress automatically. You may repeat a sound, word, or phrase across the route, and take breaths whenever needed. Plateau volume and path tolerance are adjustable; visual lift height changes the drawing, not the recorded volume reference.
Pause switches the microphone off and preserves your exact position. Resume reopens the same microphone and keeps the normal reference. A microphone or reported processing change requires recalibration. Restart flight begins the same prompt again and retains the unfinished route in the review.
An optional illustration you can hear A synthesized rising tone—not a voice, an example of correct speech, or a clinician-reviewed model. The game demo is silent.
What the trace can—and cannot—show Lift first checks an early, noise-corrected amplitude window against a fixed normal-voice reference. A three-frame causal median reduces isolated level spikes. At least 140 ms of voiced sound must be in the selected window, allowing at most 30 ms of interruption. A sustained 30 ms above the upper limit blocks that start. A beginning that cannot supply the evidence within 650 ms must be tried again. Device buffering and the point where your sound enters the window add to response time. Approximate periodicity rejects some unvoiced noise; another voice, music, or a tonal sound can still qualify. This is not speech recognition.
After takeoff, volume uses a three-frame median and a 45 ms exponential level filter. Voicing, an additional raw-level guard, and a brief settling check prevent silent or clearly out-of-window input from moving the glider. Route progress is updated only from incoming audio frames, never from the animation clock. The optional initial-rise estimate uses 10 ms frames and the beginning up to the soft-window decision. It does not control takeoff. The trace shows amplitude relative to your normal reference; 50% of amplitude is not a claim of half the perceived loudness. Microphone distance, room sound, and device processing affect these estimates. Normal volume is not silently adjusted during play. Recalibrate after changing the microphone or its position.
“Not measured” is not an incorrect attempt. Active or unreported browser audio processing disables guide comparisons. Browser-reported settings cannot rule out processing inside the microphone or operating system.
This game has not been clinically validated. Matching a soft-amplitude window does not establish that a person used gentle onset correctly. Gentle onset concerns how voicing begins, not simply quiet speech. Lift does not measure airflow, throat tension, vocal-fold contact, or treatment outcomes. A speech-language pathologist can provide individualized guidance.
ASHA: Stuttering, Cluttering, and Fluency Your audio and results Audio is analyzed in your browser. This file makes no network requests, does not use cloud speech recognition, and does not retain playable audio. Only short-lived numeric signal features and session results are held in memory. CSV export includes accepted and not-launched starts, full and partial routes, prompts, normal references, target settings, checkpoints, bounded numeric volume traces, and your optional effort rating. Nothing is saved automatically.
The microphone turns off when you finish, exit, or switch away from the tab. Returning to a paused session requires “Resume.” For hosted microphone access, use HTTPS or localhost.
Microphone permissions · Browser audio processing Ready to fly