Word removal from a song without losing the beat

Online tool for removing a word or a stray syllable from a finished song. The voice is separated from the music, the syllable is lifted out of the vocal lane only, and the freed slot is closed by pulling the rest of the phrase back — so the track never drops out and the timing stays put. The main use case is fixing a typo in an AI-generated song, where regenerating would cost you the take you liked.

How to remove a word from a song

1

Upload the song — MP3, WAV, FLAC, M4A, OGG or a video file. Decoding happens in your browser.

2

Wait for the voice and music to be separated. This step downloads the separation model, so the first run takes longer than later ones.

3

Click the syllable you want gone: on the vocal lane syllables are marked as individual blocks.

4

Pick how to close the hole, or keep the recommended option. Turn on the seam loop and listen to the splice, flipping to the original with the A/B button.

5

Download the result. The file length does not change.

Lift out a stray syllable or word — the music plays straight through

Drop audio or video here
MP3, WAV, FLAC, M4A, OGG, MP4 ·

What it does

Syllable markers instead of a word list: speech recognition collapses a repeated "na na" into one "na" and never shows the duplicate at all, so syllables are found from the 500–2500 Hz formant contour instead.
Three ways to close the hole — shift the phrase by one slot, hold the neighbouring syllable, or leave a gap — and the tool states which one it picked and why rather than deciding silently.
The music lane is locked for editing: it carries the grid, which makes breaking the timing technically impossible.
Loop over the seam and instant A/B against the original at the same playhead — a splice is judged by ear, not by looking at a waveform.
The beat grid is derived from the actual singing, not from the drums alone: syllables typically sit tens of milliseconds behind the drum phase, so a grid-derived shift would miss every time.
Outside the edited span the source audio is not recomputed — only the edited fragment changes, and the file length stays the same.

Frequently asked questions

How is this different from a censoring tool?

A censor mutes the word and leaves a hole across the whole track — music included. Here the syllable is lifted out of the voice only, and the freed time is closed musically, so there is no dropout in the music and no audible splice.

Why not just cut the piece out and join the ends?

Cutting changes the length: everything after the cut slides forward and the timing breaks. That is exactly how a song sounds after someone tries to fix it with a silence remover. Here the music lane is never cut and the beat grid is left untouched.

Why does the transcript not show the duplicated syllable?

Speech recognition is a language model: it "knows" how the line should read and collapses the repeat. On the reference file the transcript was byte-identical before and after the duplicate was removed. So syllables are found from the sound instead — from the way nasals and stops damp the 500–2500 Hz band while open vowels fill it.

Why does holding a syllable sometimes sound dragged out?

If the hole is closed by holding the neighbouring syllable, that note now occupies two slots instead of one, and the ear reads it as a delay. Shifting the rest of the phrase has no such cost, which is why it is offered first and holding is offered only when shifting is impossible.

When is shifting the phrase unavailable?

If the voice never stops after the phrase there is nowhere to shift into: every later line would move off the beat. Holding or a gap remain, and the tool says so on screen instead of quietly picking the worse option.

Does it work on AI-generated songs?

That is the main use case. The editors built into music generators regenerate the fragment, which means a different take. Here the original recording is not recreated at all — it is edited, so the voice, melody and arrangement stay exactly as they were.

Which formats are supported?

Input: MP3, WAV, FLAC, M4A, OGG and video files with an audio track. The result is exported as lossless WAV.

You May Also Like

💿 5.1 Surround Disc Image for Home Theatre 🎮 8-Bit Chiptune Synthesizer 📻 ACX Audiobook Mastering 🪄 AI Music Cleaner ⏸️ Add Silence 🔇 Audio Censor 📖 Audio Chapter Splitter 📦 Audio Compressor

How do you rate this tool?

Thank you for your rating!
Want to share more? Leave a comment!
Thank you! Your comment will appear after moderation.
Who is this tool for?
🎵 Musicians & Producers ✨ Content Creators
Published Updated Author: