Menu

SingKey

Menu

What SingKey can do today

SingKey ships nothing fake. Where a capability needs real analysis or a licensed provider that doesn't exist yet, it says so rather than inventing a result.

  • One account, every device

    Not built yet

    With an account, your singers, songs, renditions and set follow you to another phone or browser. Recordings stay on the device that made them.

    Needs a database configured for this deployment. Without one, SingKey keeps everything on this device, exactly as it always has.

  • Note and key maths

    Working

    Pitch, interval, range and transposition maths shared by every module.

  • Key recommendation engine

    Working

    Scores all twelve keys against a singer profile and ranks them.

  • Pitch detection

    Working

    SingKey hears which note you are singing, on your device — nothing is uploaded. Measured against generated signals, never yet against a real voice (docs/VOICE_BENCHMARK.md).

  • Guided voice check

    Working

    Hear a phrase, sing it back, and SingKey works outward from the middle of your voice.

  • Human demonstrations

    Not built yet

    A real singer showing each exercise, with a posture and breathing reference.

    Needs recorded or licensed footage of a real singer. SingKey will not substitute a cartoon, stock illustration or animated effect, so exercises use a labelled reference tone until then.

  • 5-minute warm-up

    Working

    A seven-exercise guided routine. Not reviewed by a vocal professional yet, and it says so.

  • Key detection

    Working

    Finds the key of a recording with chroma analysis. Benchmarked in docs/SONG_BENCHMARK.md.

  • Melody analysis (single line)

    Working

    Measures the notes, important range and demanding phrases of one sung or played line. Benchmarked in docs/SONG_BENCHMARK.md.

  • Sing a song to find your key

    Working

    Sing the part you normally sing and SingKey measures the melody in your voice, on your device. Nothing is recorded or uploaded.

  • Hear your melody in another key

    Working

    Plays back the melody you sang, moved into the key being recommended, so you can hear whether it fits before committing. Rendered as a tone — your melody, not your voice.

  • Sing along to an accompaniment

    Working

    Chords played in your chosen key, synthesised from the public-domain chord charts SingKey holds. Not a recording of anyone.

  • Lyrics that follow the music

    Working

    The words move with the accompaniment. Exact rather than estimated, because SingKey generates the accompaniment and so defines the timeline.

  • Metronome

    Working

    A count from 30 to 300, scheduled against the audio clock so it does not drift.

  • Guitar tuner

    Working

    Six strings in standard tuning, read by the same pitch detector the voice screens use.

  • Bass tuner

    Working

    Four strings in standard tuning, read by the same pitch detector.

  • Lyrics that follow a recording

    Not built yet

    Words timed to an original recording rather than to SingKey’s own accompaniment.

    Following a recording needs alignment data nobody has, and timing driven by elapsed time would drift and mislead. Sing-along works because SingKey generates the accompaniment itself.

  • A voice talking you through it

    Working

    A generated speaking voice reads each instruction while reference tones sound the pitch. It is speech, not singing, and never presented as a person. Where no speech service is configured the words are shown as text and the tones still sound — so this is listed as available because the exercise always works, not because a voice is guaranteed.

  • Hear a phrase sung

    Working

    A synthesised singing voice sings an original phrase in your key — real sustained pitches and vowels, rendered on your device. Verified as singing rather than speech by SingKey’s own measurement.

  • A real person singing the reference

    Not built yet

    Hearing a human sing the phrase before you sing it yourself.

    No recording of a real person singing has been obtained with licensing clear enough to use. Until one is, SingKey plays a guide chord in the song’s key and labels it a reference tone — never a human, never the tune.

  • Following the words as you sing

    Not built yet

    Highlighting the line you are currently singing.

    Following a voice through a lyric needs alignment data these songs do not have. A highlight driven by elapsed time would drift and then mislead, so the words are shown as a prompt instead.

  • Meet your voice by just singing

    Working

    Sing or hum anything for about a minute and SingKey measures where your voice sits. No notes to match.

  • Where your voice sits

    Working

    A hedged likeness — "most similar to a baritone range" — drawn from your comfortable range, never from one reached note. Never a classification, and "not enough evidence yet" is a normal answer.

  • Speaking voice measurement

    Not built yet

    Measures where your speaking voice sits, shown separately from your singing range and never used in place of it.

    The measurement itself is built and tested, but no screen a singer can reach offers it yet — today it only runs in the developer-only Test Lab, which does not exist in a production build.

  • Your recordings, on this device

    Working

    Keep a take, hear it back, delete it. Stored in this browser and nowhere else — nothing is uploaded, and deleting one removes the audio for good.

  • Several singers on one phone

    Working

    Each person gets their own name and their own measurement, so a household or a worship team does not overwrite each other.

  • Add your own song

    Working

    Give a title, the original key and roughly where the tune sits — plus artist, tempo, words and chords if you have them — and SingKey answers for that song. Keep it and you can come back to it, sing along to it, and edit or delete it. It stays on this device: a song saved on one phone is not on another.

  • Watching your voice as you sing

    Working

    Your voice drawn live against the range you sang comfortably, with a musical grid and the reference path where there is one. Gold is you. It breaks where you breathe rather than drawing a note you never sang.

  • Choosing the key before you sing

    Working

    Best fit for your voice, the song’s own key, or another option — picked and heard before the microphone opens. It never starts in a key you did not choose.

  • Do re mi

    Working

    Five notes played on the piano, then sung back, three times — low, middle and a little higher. Your voice is drawn over the notes you were asked for. Nothing is scored, and it never climbs until you fail.

  • Seeing where your voice lives

    Working

    Your comfortable singing area drawn as a shape rather than written as two note names. Lit gold where you sing comfortably, dimmer where SingKey has merely heard you. No note names unless you ask.

  • Opening a recording

    Working

    A saved take shows what you sang, your comfortable area, the key SingKey would suggest from it and why — not just a play button. Facts it never recorded are left out rather than guessed.

  • What happened last time

    Working

    Come back to a song and SingKey tells you which keys you have tried and how the last one felt. Never treated as today’s answer — your voice is different on a different day, so it offers to test again.

  • Comparing two keys you sang

    Working

    Sing the same song in two keys and see both. SingKey says which one it heard more comfortable singing in — and asks you which felt better, because that is the part it cannot measure.

  • Your voice, learned over time

    Working

    Takes recorded on a song screen are kept as evidence. With an account, that evidence follows you to another device.

  • Compare this take with your last

    Working

    What changed between two recordings of the same song.

  • Chord chart in your key

    Working

    The chart for a built-in hymn, transposed to the key you chose, with a simplified view and Focus Mode.

  • Verified real-world accuracy

    Not built yet

    How well the analysis performs on real recordings of real singers.

    Every accuracy figure SingKey publishes comes from synthetic fixtures. No real recording has been through the corpus harness yet, so no real-world claim is supported (docs/research/VALIDATION_CORPUS.md).

  • Melody analysis from a full mix

    Not built yet

    Extracting a lead vocal's melody from a produced recording.

    Separating a voice from a band needs source separation this deployment cannot carry, so analysis refuses at the polyphony gate rather than transcribing the piano as your melody (ADR-0031).

  • Lyrics

    Not built yet

    Lyrics for songs beyond the public-domain hymns SingKey ships. The built-in hymns already show their own words; this is about everything else.

    No lyrics provider is configured. Paste a link or the words on the import screen.

  • Chords

    Not built yet

    Chord charts for songs beyond the public-domain hymns SingKey ships, transposed to your key. The built-in hymns already have charts.

    No chords provider is configured. Paste a link or the chart on the import screen.

  • Your set

    Working

    Save a song, the key you’ll sing it in and how the hard part sat, for the people preparing the service. Kept on this device; not shared yet.

  • Practise the hard part

    Working

    Sing the hardest part of a song you sang, with your own melody drawn and short feedback. Judged by SingKey’s measurement, never yet by a person.