An independent audio tool designed to identify background music in videos when speech, commentary, or background sounds stop standard apps from matching.
Why Standard Song Recognizers Miss Video Tracks
Traditional music recognition apps listen to the entire audio waveform at once. When a video contains a creator speaking, narration, sound effects, or audience chatter, those foreground voices corrupt the acoustic fingerprint.
BGM Finder fixes this by running an AI vocal separator before running catalog searches. It strips human dialogue and leaves the background instrumental clean for acoustic fingerprinting. Human speech primarily occupies frequencies between 300 Hz and 3,400 Hz, directly overlapping the melodic overtones, chord progressions, and synthesizer leads that define a song's musical identity. By isolating stems first, recognition engines can match the clean backing track without voice interference.
The Step-by-Step Audio Pipeline
Here is what happens behind the scenes from the moment you submit a link or file.
1
Audio Extraction & Normalization
Audio is pulled directly from the submitted link or uploaded file, decoded to high-quality audio in memory, and normalized for recognition.
2
Neural Vocal Separation
If speech obscures the track, an open-source neural separation model (UVR5 MDX-Net ONNX) performs spectrogram phase-aware masking to isolate and suppress spoken voice frequencies while preserving melodic lines, basslines, percussion, and background harmony.
3
Shazam & Description Music Match
The isolated instrumental stem is transformed into an acoustic landmark constellation map. These acoustic hashes are cross-referenced against Shazam's music catalog, alongside checking YouTube's music credits in the video description to verify the song title, artist, album, and release information.
4
Results & In-Browser Preview
View verified song metadata with direct links to Spotify, Apple Music, and YouTube. For uploaded audio or video files, preview or download the 15-second isolated instrumental stem directly in your browser.
Supported Platforms & File Formats
What you can paste or upload into BGM Finder.
YouTube & Shorts
Standard links, youtu.be, Shorts, and streams
Supports timestamped links (?t=1m30s) and includes an interactive segment picker for videos longer than 2 minutes.
TikTok Videos
Web links, mobile app shares, and audio pages
Paste direct links. If creator speech covers the music, the separator isolates the underlying backing track.
Instagram Reels
Reels, feed video posts, and Stories
Supports reels links. Useful for finding songs in workout videos, cooking tutorials, and vlogs where speech is constant.
Direct File Uploads
MP3, WAV, MP4, M4A, AAC, WebM, FLAC (up to 25MB)
Drop screen recordings, voice memos, or downloaded video clips directly into the browser dropzone.
Privacy & Ephemeral Processing
No accounts, no permanent storage, no tracking
Audio clips are processed entirely in temporary storage (in memory when available, otherwise on disk) and deleted automatically within 30 minutes of search completion. I do not store user files or distribute copyrighted tracks.
Technical Boundaries & Limitations
What our neural matching pipeline can and cannot identify
BGM Finder is designed for commercial and released music. It cannot identify unreleased bootlegs, spontaneous acoustic jams, custom video game audio that lacks commercial catalog distribution, or tracks masked by extreme non-vocal audio distortion (such as continuous heavy engine roar, gunfire, or loud crowd screaming). Furthermore, video links must be publicly viewable without requiring user login or account authentication.
Who Runs This
Developer & Maintainer
Hi, I'm Srajan, an independent developer based in India. I built BGM Finder to solve a personal frustration: identifying background music in video clips when speech, narration, and background noise overwhelm standard recognition tools. Combining neural vocal separation with acoustic fingerprinting, it isolates backing tracks so creators and music enthusiasts can discover songs without friction.
Have questions, suggestions, or feedback? Reach out directly at [email protected].
Ready to Identify That Song?
Paste a video link or drop a file on the homepage to find the background track.