also searched as: identify song by humming · hum a beat to find a song · hum a song to find it · hum to search · hum to find song · search song by humming
Hum, whistle or sing 10 to 15 seconds of the tune into a song search and AI names it, with a short ranked list of likely matches. Off-key is fine: the model keeps the shape of the melody and throws away your voice. It is free in the search and music apps listed below, on Android and iPhone, though on an iPhone you install one first.
checked 2026-10-13 · by dan scutaru
the tools that earn the verdict — no pay-to-win
Open a song search, hum for 10 to 15 seconds, then pick from the matches. In the search app, tap the microphone and choose the song search button; in the music apps listed above, look for a waveform icon or a sing option next to the mic. Hum the chorus or the hook, the part you would sing in the shower. You get a short list of likely songs, and tapping one plays it.
If the first try misses, try a different part, or sing it on "da da da" instead of a closed-mouth hum: a hum carries little information, and clear note changes give the model more to match. Perfect pitch is not needed, since the system is trained to cope with a different key and tempo.
The match runs on melody alone. The model turns your hum into a sequence of numbers that traces the tune going up and down, and strips out everything else: instruments, backing vocals, the sound of your voice. A song's melody works like its fingerprint, which is why a shaky, off-key hum can still land on the right track.
That also explains the misses. A drum pattern, a beatboxed rhythm or one repeated note has no shape to compare, so hum whatever carried the tune over that beat. The reference catalog is built from the original recordings, so recent releases can be matched without anyone having to hum them first.
When every search comes back empty, the song is often missing from the catalog rather than lost in your humming. The only published catalog size for hum search, from its launch, was over half a million songs, while a streaming service lists tens of millions of tracks. Popular songs with a clear tune match best; obscure tracks, local releases and instrumental scores are where results get thin.
Try a second tool from the list, since they do not all run on the same engine. If the tune came from an ad, a show or a game, search for that instead, with the word "song" after it. Last resort: record yourself humming and post it to a song identification forum, where people listen instead of a model.
Every tool listed here does hum search in a phone app, and none charges for it. On Android, asking the phone's assistant "what song is this?" also opens a screen that takes a hum. On an iPhone you install an app first; the search app's song button works the same way there. Two of the picks are streaming apps, so you can play or save the match straight away. The dedicated recognition app shows ads unless you pay once to remove them.
A computer is harder. The best-known website that took a hum in the browser now forwards you to a phone app download page, so on a laptop the quickest route is the phone in your pocket.
Yes. Ten to fifteen seconds of the melody into a song search usually brings back the title, ranked against a few close guesses. It works best on well-known songs with a clear tune, and it costs nothing in any of the apps listed above.
Singing usually gives the best results. The engine behind hum search worked well for singing first and needed extra training before it handled humming and whistling, because a closed-mouth hum carries little information. If you do not know the words, sing the tune on "la" or "da"; whistling works too.
Yes, with an app. All four picks above run on iPhone: the search app's song button, the waveform search in one music app, the sing mode in the other, and the dedicated recognition app. Install one, allow the microphone, and hum for 10 to 15 seconds.
Only the ones with a hum or sing mode. An app built to name a song playing near you matches the original recording, and your hum is not that recording, so it usually comes back empty. Use one of the tools above for humming; for a track that is actually playing, see whether a chatbot can name it for you.
Rarely. Matching runs on melody, the rise and fall of notes, and a beat has none to compare. Hum the vocal line or the riff that sat on top of that beat, or type any words you remember into an ordinary search.