Which file do I download to get the voice without music?

Download the vocals stem — that is the isolated voice with the background music removed. The same run also gives you the instrumental, in case you want the music on its own.

Does it work on speech, podcasts, and videos, not just songs?

It works best on a clear voice over music — songs, covers, streams, and voiceovers with a music bed separate very cleanly. Spoken podcasts with a subtle bed also work, but a very quiet or heavily-processed bed may leave faint traces, since the model is tuned for vocals over music.

Can I keep the background music instead?

Yes — the same separation also produces the instrumental (music-only) stem, so you can keep whichever side you need, or both.

What format should I download?

WAV or FLAC for editing and re-use — lossless keeps the voice natural. MP3 320 is fine for quick listening or a phone.

Free AI to Remove Background Music — Keep Just the Voice

Pull the voice out of any song, cover, stream, or recording and leave the music behind. A studio-grade three-model ensemble separates the vocal from the backing track, so you can download a clean voice in lossless WAV or FLAC — from a file or straight from a link.

🎙️ Keep the voice 🎚️ Lossless WAV / FLAC 🔗 File or URL

Remove the music in about 6 minutes

🎙️ Open the music remover

Free · download the vocals stem · no signup

How to remove background music — 4 steps

Upload the audio or paste a YouTube / SoundCloud / TikTok URL.
Pick WAV or FLAC. Lossless keeps the voice natural if you plan to edit or re-use it.
Click Separate. The Studio pipeline runs three transformer-based models and averages their masks to split the voice from the music.
Download the vocals stem — that is your voice with the background music removed. (The instrumental stem is there too if you want the music alone.)

What separates the voice cleanly from music

Removing background music well means three things: the voice stays natural (no robotic or underwater artifacts), the music actually leaves (no drums or melody bleeding under the voice), and nothing important is lost (breaths, consonants, and tone intact). Old "music removers" used phase tricks that wrecked all three.

Modern AI models like BS-Roformer learn what a voice looks like in the spectrogram and keep only that, discarding the rest as music. Our ensemble blends three top models so no single model's weakness — one leaking cymbals, another dulling the voice — shows up in your result.

Who removes background music

🎬 Video creators

Isolate a voice from a clip so you can drop in your own music, or clean up a track that shipped with an unwanted bed.

🎙️ Podcasters & voiceover

Recover a clean voice from a recording that has a music bed baked in, ready to re-mix or re-caption.

🎤 Cover & vocal reuse

Get the pure vocal from a song or cover to study, re-time, or layer over a new arrangement.

🈶 Dubbing & language work

Pull the spoken or sung voice off the music before transcribing, translating, or re-dubbing.

📞 Interviews & lectures

Lift a voice out of a noisy, music-laden recording so the words are clear and easy to caption.

🎧 Sampling & remix

Producers grab a clean vocal to chop, pitch, and rearrange into something new.

Why our music removal beats other free tools

Three-model ensemble vs a single model. Most free tools run one older model that either leaves music bleeding under the voice or makes the voice sound thin. Averaging three transformer models keeps the voice clean and full.
You get both sides. One run gives the voice-only stem and the music-only stem, so you keep whichever you need.
Lossless WAV / FLAC. Many free tools export only MP3, which softens the voice. We give you the full-quality stem for editing.
Straight from a link. Paste a YouTube, SoundCloud, or TikTok URL — no separate download step.
Private & auto-deleted. Files run on our own GPU and are removed within 24 hours.

Frequently asked questions

Which download is the voice with no music?

The vocals stem. It is the isolated voice with the background music removed. The instrumental stem is the opposite — music with the voice removed.

Will it work on a podcast or a talking video?

It works best on a clear voice over music (songs, covers, streams, voiceovers). Spoken content with a music bed usually cleans up well too, but a very faint or heavily-processed bed may leave slight traces — the model is tuned for voice over music, not arbitrary noise removal.

Can I keep the music and remove the voice instead?

Yes — download the instrumental stem from the same run. See the instrumental extractor for that use case.

How long and how big can the file be?

Up to 20 minutes per file, up to 100 MB. A typical song finishes in about 6 minutes.

Does it change the voice or add effects?

No — it only separates. The voice keeps its natural tone; the output is loudness-normalized but not otherwise processed.

Free tier: how many files?

3 songs per month free once you sign in (1 free to try without an account), or more with Patreon Pro or pay-as-you-go credits.

Related free tools

Keep the voice, remove the music

🎙️ Open the music remover

voice-only stem · WAV/FLAC · free · no signup