Vocal Isolator Review: Vocals and Background in One Call
Pick Vocal Isolator for rehearsal tracks and cleaning up narration; listen to both stems, because leakage between them is normal for this kind of tool.
Vocal Isolator takes one recording through audio_url and returns two related files from the same job: the vocals first, then everything else as the background. In a song the background is the accompaniment; in a spoken clip it can include music and room sound. There is no prompt, no speaker selector and no output setting. Recordings from 1 to 600 seconds are accepted, and the price is $0.0012 per input second, rounded up, so a three-minute song costs $0.216 and getting two stems does not double it.
- SpicyAPI Engineering
- Tested Oct 24, 2026
The verdict
It is a utility for sound you already have: a backing track to sing over, a clip of speech over music you want to rework, a podcast with a bed underneath. Keep the original mix alongside the stems so you can always compare against it.
Key facts
- Content ladder: not applicable; this endpoint separates a recording you supply.
- Price: $0.0012 per input second, rounded up.
- Input:
audio_urlonly, 1 to 600 seconds, MP3 or WAV through the upload control. - Output: two files from one request, vocals and background.
- Sources: the parameter list is on the Vocal Isolator model page.
What we liked and what held it back
What we liked
- One field.
audio_urlis the entire request. - It is very cheap. A 30-second clip costs $0.036 and a full 10-minute file $0.72.
- Both stems come from one pass. They are two parts of the same recording, meant to be auditioned together or used in a new mix.
- Price follows the source length. Nothing about the output changes the bill.
- Ordinary audio files came back. Our first run returned two 30-second MP3 files at 44.1 kHz stereo, the same length as the input.
What held it back
- Separation is never perfect. Reverb, backing vocals and quiet details that overlap between the two parts can leak into the wrong stem or be lost.
- No speaker control. You cannot ask for one voice out of several, and there is no prompt to steer it.
- The background of a spoken clip is not just music. It can hold room sound and any other non-voice audio.
- Clean sources help. Send the clearest version you have, not a re-recording of loudspeaker playback or an extra-compressed copy.
- Ten minutes is the cap. Cut longer recordings before sending.
What we tested
We have one real run, from 24 October 2026: a 30-second mix of a voice over music. It returned two stems, both 30 seconds long, in MP3 at 44.1 kHz stereo, billed on the 30 seconds of input.
We confirmed the delivery and the format, not the separation quality. Nobody has yet scored how cleanly the vocals sit apart, how much music bleeds into the vocal stem, how the two stems line up against each other, or how it behaves on speech-only recordings. Treat those as open questions and listen through pauses and quiet passages on your own file before using either stem.
Alternatives
- Dubbing a clip into another language? Video Translator re-voices the speech and can keep or drop the background under it.
- Need sound for a silent clip instead? Video Sound Effects generates ambience and effects from the picture.
What Vocal Isolator can do on SpicyAPI
Straight from the live model specification on SpicyAPI. Some creators offer more in their own apps; this is what you can call through our API today.
| Feature | audio-separation |
|---|---|
| Seed control | No |
| Negative prompt | No |
What do you think?
Was this review useful?
Log in to vote.
Comments
Comments are checked automatically and by our team. We remove spam, offensive content, personal attacks and anything that names a real person without consent. Report a comment and we review it within one business day.
Tried this model? Share what worked. You need a free account to comment, so we can keep spam out.
Vocal Isolator FAQ
Try Vocal Isolator yourself
Set every parameter in the Playground and see the exact price before you run it.