Find Song by Audio: Upload a Clip, Get the Name

Upload the file and let the match come back to you. Song Finder accepts audio in mp3, wav, flac, m4a, aac, ogg, and webm, plus mp4 and mov video, and reads the file directly instead of listening to it played out loud. Candidates return with cover art, artist, album, genre, and a confidence score, and an on-page preview confirms the winner. That's the whole find song by audio flow: drop the file in, check the shortlist, follow the link out.

Audio uploads can be mp3, wav, flac, m4a, aac, ogg, or webm; video uploads can be mp4 or mov. Each file can be up to 40MB, which comfortably fits full songs, long voice memos, and multi-minute clips. If a recording session produced something larger, trim it to the section where the music is clearest and upload that part; the search needs one clear stretch of sound, not the entire event.

Often, yes. Because Song Finder reads the uploaded file directly, nothing further is lost to speakers, room echo, or a second microphone, so even a rough voice memo keeps whatever signal it originally captured. The confidence score then tells you how solid the result is: a high score settles it, and a borderline one is your cue to try a longer or cleaner section of the same recording and compare the candidates again.

Yes. An mp4 or mov file is searched by its soundtrack: Song Finder extracts the audio internally and runs the same catalog match, so you never pull the audio out yourself. It's the simplest way to learn what is this track inside a clip saved on your device, whether that's a concert video, a screen recording, or something forwarded from a group chat.

No. Song Finder is a song identifier online, an application to recognize music that runs completely in your browser on phones, tablets, and computers. Uploads work through the normal file picker, nothing is installed, and there is never an update to manage. That matters most on work machines and shared devices, where adding apps isn't always an option.

It tells you how certain the match is. A score of 70 or above is a confident match, while results around 40 to 70 are low-confidence candidates, kept visible because covers, remixes, and rough recordings sometimes land there while still being right. Low-confidence results also stay in your history, so you can search a song again later with a better clip and see whether the score climbs.

A no-match is an honest outcome with three usual causes: the song isn't in the catalog, the audio is too noisy or too quiet, or it's an AI-generated track that hasn't been indexed. The attempt is saved to your history, so retry with a longer or cleaner stretch of the file. If the music also exists online, pasting its link is a second route; sometimes the platform's copy is cleaner than yours, and that second pass is what manages to id the song.

Yes, that's the record mode's job rather than upload's: it answers what song is playing through your microphone or system audio when there is no file to begin with. Humming covers the case where the music lives only in your memory, and link pasting covers audio that lives online. Upload stays the best route whenever the sound already exists as a file on your device.

Fresh enough for recent releases to be realistic matches. New tracks join the catalog daily by the tens of thousands, on top of a base well past the hundred-million mark, so an unlabeled rip of something that came out this month is still worth uploading. If it misses today, the retry you run from history next week may land, because the catalog will have grown in between.

Results land in a history stored in your own browser, 20 entries deep, matched and unmatched alike. You can filter them, reopen a result instead of re-uploading, delete entries one by one, or clear everything behind a confirmation step. Because the list lives locally, it stays on the device you worked from, which suits the upload workflow of cleaning one folder on one machine.