AI Vocal Remover
Ask fal-ai/sam-audio/separate for vocals or another described sound and receive that target plus the remainder. Dense mixes can produce bleed or artifacts, so review both stems.
Drop an audio file here
or click to browse
How it works
Load a song
Drop in any track. WAV or FLAC gives the cleanest separation; MP3 works great too.
Pick what to isolate
Choose "vocals" or describe a distinguishable sound such as drums, bass, guitar, or speech. The model attempts to split that target from the remainder.
Download both stems
Preview each track on the page and download the ones you want. Workspace access to encrypted retained server objects expires on your plan’s schedule. Expired objects are queued for automatic cleanup, and deletion is retried if storage removal fails.
Frequently asked questions
Do I get both the vocals and the instrumental?
A successful fal-ai/sam-audio/separate job returns the requested sound as one track and the remainder as another. Bleed and artifacts are possible, especially in dense or heavily processed mixes.
Can I isolate sounds other than vocals?
You can prompt for sounds such as drums, bass, guitar, speech, or birdsong. The model attempts that separation, but success depends on whether the sound is distinguishable in the mix.
What quality can I expect?
Quality varies by mix, source compression, and prompt. Dense arrangements, effects, and overlapping frequencies can cause bleed or artifacts; lossless WAV or FLAC avoids adding another lossy source.
How much does it cost?
Stem separation costs 3 credits per minute of audio. A free account includes monthly AI audio credits to try it — no card required — and Pro plans include 300 credits per month.
Cleaning up a voice recording instead? Try AI noise removal or the full Audio Studio.