Voice isolator

Voice isolator features

• 01

  Two-stage isolation, not just denoise

  Multi-pass neural-network denoising tightens the noise floor on each pass. A voice-activity-driven gate then silences frames the model is confident contain no speech — so background music, applause, and conversation drop out entirely between phrases.

• 02

  Tunable for podcast or rescue work

  Isolation strength controls how aggressively non-voice frames are gated. Lower it for natural-sounding podcasts, push it up to fully strip a music bed or crowd from a noisy recording.

• 03

  Runs locally, no upload

  The whole pipeline is a 125 KB WebAssembly module that loads once and stays cached. Audio is decoded, isolated, and downloaded entirely on your machine — no server round-trip, no account, no length limits beyond the 200 MB file cap.

Voice isolator FAQ

How is this different from the Voice Enhancer?

Voice Enhancer runs a single denoise pass and keeps the natural feel of the recording — best for cleaning up steady noise like fans or AC hum. Voice Isolator stacks multiple passes and adds a voice-activity gate that silences anything outside the speaker — best for stripping music, crowds, or another conversation from the background.

Can it remove background music from a voice recording?

Yes, when the music is clearly behind the voice in level. The gate silences frames with no detected speech, and the multi-pass denoise pulls down music bleeding through during words. Heavy mastered music at the same loudness as the voice is harder — try strength 90–100 and three passes for those cases.

What does the isolation strength slider do?

It sets how aggressively non-voice frames are attenuated. At 0 the gate is loose and you'll hear faint background; at 100 anything the model isn't confident is voice goes to silence. 70–80 is a good starting point for podcasts, 90–100 for music or crowd removal.

Is the audio uploaded to a server?

No. The model and your file stay in your browser. The pipeline is a small WebAssembly module that runs locally on your CPU, so nothing leaves your computer.

What output format do I get?

A mono 48 kHz WAV file in 16-bit PCM. WAV is uncompressed and works in every audio editor and podcast host. Use the Convert audio tool to export an MP3 if you need a smaller file.

How long can the recording be?

Files up to 200 MB are accepted. Two passes process at roughly 3–5× real-time on a modern laptop, so a 10-minute recording isolates in two to three minutes.