Remove the noise, keep only the voice you want

Runs entirely in your browser. Your audio never leaves your device.

Model

With a single voice plus noise, ailia Voice Filter on the “Noise Reduction” tab performs about the same as DeepFilterNet3. With multiple voices plus noise, only ailia Voice Filter — which can pick a speaker — works. Switching models takes a few seconds.

Play through speakers

Add noise to test with

Press the button to start

Before

After

Latency — Model — Processing On-device, nothing sent

If you use the microphone, your browser will ask for permission the first time. HTTPS is required except on localhost.
With “File”, the chosen audio is processed while it plays back in real time. Both Before and After can be played back and saved as WAV.
“Speaker Extraction” is supported by ailia Voice Filter only. Register a few seconds of audio where only that person is speaking, and everyone else’s voice is treated as noise too.

Remove the noise, keep only the voice you want

1Teach it your voice

Learned voice

2Try it

Before

After