Use Case
Turn YouTube MP3s into Clean Instrumentals
Already have an MP3 from a YouTube source? Use Opus to strip vocals, keep music, and create tracks for karaoke practice, remixes, and more.
Drop your MP3 here
or click to browse · max 500 MB
No account needed · File deleted after 1 hour
YouTube is one of the most common sources for MP3 files — millions of songs and recordings are available nowhere else, from rare live sessions to regional releases that never got a digital store distribution deal. When you have an MP3 ripped from YouTube and want to remove the vocals, the process is straightforward: upload it, and the AI handles the rest regardless of where the file came from.
The main consideration with YouTube-sourced files is audio quality. YouTube re-encodes audio at various bitrates depending on the upload. Most music videos deliver audio at 128–192 kbps AAC, which is decent but not lossless. Converting that to MP3 introduces a second compression pass that degrades quality further. For casual karaoke practice, this is fine. For anything you want to produce professionally, start from the highest quality source you can find.
A lower-quality source also affects how well the AI separates the vocal. The Demucs model works from the audio signal — if that signal contains compression artifacts, the separation has to work around them. In practice, 128 kbps AAC-sourced MP3s still produce usable instrumentals. The vocal removal will work; there may just be slightly more residual artefacts than you would get from a high-quality studio master.
One advantage of YouTube-sourced audio is that live performances, acoustic sessions, and cover versions are all separable. If a song was only ever recorded in a live format, AI removal is the only way to approximate an instrumental. Results vary by room acoustics and crowd noise, but isolated acoustic performances often separate surprisingly well.
For songs available in both a YouTube rip and a proper digital download, it is worth comparing both. The better the source, the better the output. But if a YouTube MP3 is all you have, the tool will work with it — processing logic is the same regardless of where the file originated.
Tips
How to get the best results
Aim for the highest bitrate available
When downloading from YouTube, use a tool that gives you 320 kbps or the original AAC stream rather than a low-quality re-encode. Better input means better separation.
Avoid files below 128 kbps
Very low bitrate files have significant frequency information missing. The AI can still process them, but expect more artefacts in the output than you would get from a cleaner source.
Check if a better source exists
Before processing a YouTube rip, check Bandcamp, SoundCloud, or a streaming service. A higher-quality download from an official source will almost always produce a cleaner instrumental.
Live recordings can separate well
Acoustic live sessions with just voice and guitar often separate cleanly — the sparse arrangement gives the AI a clear signal. Noisy crowd recordings are harder but still worth trying.
Convert to MP3 before uploading if needed
If your download tool gives you an M4A or WebM file, convert it to MP3 first. The tool accepts MP3 files up to 500 MB.
Also used for
FAQ
Frequently asked questions
Does the audio source (YouTube vs streaming) affect separation quality?+
Yes, but usually not dramatically. Lower-quality sources produce slightly more vocal bleed-through and artefacts. For practice purposes, YouTube-ripped MP3s are more than good enough.
What bitrate should my MP3 be?+
192 kbps or above gives consistently good results. 128 kbps works for practice but may have more artefacts. Below 128 kbps quality degrades noticeably.
Is processing a YouTube rip any different from processing a store download?+
No — the tool processes whatever MP3 you upload. The AI does not know or care where the file came from. Source quality affects output quality, but the workflow is identical.
Can I process videos that are music performances, not just official songs?+
Yes. Live performances, acoustic sessions, covers, and original recordings all work. Sparse arrangements (voice and instrument) often separate better than dense studio mixes.
Is there a file size limit?+
Files up to 500 MB are accepted. Most songs, even at high bitrates, are well under this limit.