Gate 0 · Fixture recording
Two lossless streams per take (AudioWorklet at 48 kHz + MediaRecorder PCM at the mic’s native rate), processing requested off and then verified from the signal. Takes are kept on this device and uploaded to your private cloud inbox; the laptop archives them into fixtures/recordings/. Protocol fixtures-v1.
- Same room, same spot, same time of day for both devices if you can. Quiet room, no fan or TV.
- Phone: hold it upright, mic end about a hand-span (~20 cm) from your mouth, the same way for every take. Laptop: sit at your normal distance and don’t move between takes.
- Before the first take: speak a few words and check the meter never hits the red. If it does, back off a little.
- Do the steps in order. The processing-ON control run is the last step on purpose.
Setup
Cloud inbox
Storage on this device
Persistent: undefined · used n/a MB of n/a MB · 0 takes on this device (0 uploaded)
Capture-chain self-test (processing OFF)
Unplug headphones and disconnect Bluetooth. Put the device on a table, volume around 60%, and stay silent for 25 s while it plays tones, noise and sweeps. This checks from the signal that echo cancellation, noise suppression and auto-gain are really off.
Route: built-in speaker / no headphones · 26 s
Room silence
Hold the device where you will hold it when singing. Stay completely silent for 10 s.
Route: built-in speaker / no headphones · auto-stops after 10 s
Sustained “ah”: lowest comfortable note
Pick a lowest comfortable note (the reference is optional: pick a note, play it, then match it). When the screen says SING, sing “ah” steadily for about 5 s at a moderate volume, about the same loudness on every vowel take. Don’t push.
Route: built-in speaker / no headphones · auto-stops after 7 s
Sustained “ah”: low-middle note
Pick a low-middle note (the reference is optional: pick a note, play it, then match it). When the screen says SING, sing “ah” steadily for about 5 s at a moderate volume, about the same loudness on every vowel take. Don’t push.
Route: built-in speaker / no headphones · auto-stops after 7 s
Sustained “ah”: middle note
Pick a middle note (the reference is optional: pick a note, play it, then match it). When the screen says SING, sing “ah” steadily for about 5 s at a moderate volume, about the same loudness on every vowel take. Don’t push.
Route: built-in speaker / no headphones · auto-stops after 7 s
Sustained “ah”: upper-middle note
Pick a upper-middle note (the reference is optional: pick a note, play it, then match it). When the screen says SING, sing “ah” steadily for about 5 s at a moderate volume, about the same loudness on every vowel take. Don’t push.
Route: built-in speaker / no headphones · auto-stops after 7 s
Sustained “ah”: highest comfortable note
Pick a highest comfortable note (the reference is optional: pick a note, play it, then match it). When the screen says SING, sing “ah” steadily for about 5 s at a moderate volume, about the same loudness on every vowel take. Don’t push.
Route: built-in speaker / no headphones · auto-stops after 7 s
Sustained “ee”: middle note
Same middle note as before, on “ee”, about 5 s.
Route: built-in speaker / no headphones · auto-stops after 7 s
5-note scale, low start
The app plays do-re-mi-fa-sol-fa-mi-re-do once. Then, during recording only, sing it back on “ah”, about one note per half-second.
Route: built-in speaker / no headphones · auto-stops after 9 s
5-note scale, middle start
Same, starting higher.
Route: built-in speaker / no headphones · auto-stops after 9 s
Siren 1
On “oo” (or “ng”), glide gently from low to high and back down, about 3 s up and 3 s down. Only go as high as stays easy.
Route: built-in speaker / no headphones · auto-stops after 9 s
Siren 2
Repeat the same siren.
Route: built-in speaker / no headphones · auto-stops after 9 s
One verse, a cappella
Sing one verse of a song you know well, unaccompanied. Remember which song and verse: you’ll sing the same one next. Optional: play your starting note first. Stop when done.
Route: built-in speaker / no headphones · stop manually
Same verse, headphones + backing track
Plug in WIRED headphones if you have them (note it below if they’re Bluetooth). Load a backing-track file you have the right to use. Stay SILENT for the first 5 s (this measures how much of the track leaks into the mic), then sing the same verse. Stop when done.
Route: headphones · stop manually
Control run (processing ON, expected to FAIL)
Do this LAST. Headphones out, same setup as the first self-test. This deliberately turns the browser’s voice processing ON. If the test still passes, it is not sensitive on this device and the first result proves nothing. (It runs last because on iOS a processing-ON mic can switch the shared mic unit into voice-processing mode.)
Route: built-in speaker / no headphones · 26 s
Gate 1 · device tests
Latency: built-in speaker
Headphones out, Bluetooth off, volume ~60%, stay silent. Five short sweeps; measures the round trip speaker to mic.
Route: built-in speaker / no headphones · ~8 s
Latency: wired headphones (skip if none)
Plug in wired headphones and set Output route above to "wired headphones". Hold one earbud right against the device mic (phone: bottom edge; laptop: near the built-in mic) during the test.
Route: headphones · ~8 s
Latency + quality: Bluetooth headphones (skip if none)
Connect Bluetooth headphones/AirPods, set Output route to "Bluetooth headphones", hold an earbud against the device mic. This also records what Bluetooth does to the mic signal.
Route: headphones · ~8 s
Through-the-mic replay (this device hears itself)
Headphones out, volume ~60%, quiet room, stay silent ~55 s. Plays tones with exactly known pitch through the speaker and records them through the mic.
Route: built-in speaker / no headphones · ~55 s
Through-the-mic replay from the other device: RECORD here
Phone only. Put the phone ~50 cm from the laptop speakers. Press Record here first, then press Play stimulus on the laptop, wait until it finishes (~55 s), then Stop here.
Route: built-in speaker / no headphones · stop manually
Through-the-mic replay from the other device: PLAY here
Laptop only. Plays the stimulus through the laptop speakers for the phone to record (nothing is recorded here).
Route: built-in speaker / no headphones · ~55 s