Clean voice recordings for podcasts, video, interviews and more

Four steps, whatever you record

Every use case goes through the same short flow in Noise Remover. What changes from one to the next is the way you record, the settings that matter and the limits to watch for.

1. Bring the audio in

   Record in the app, import an audio file, pick a video, or share a file from another app: a voice note, a recorder, your gallery. Every format it opens is listed on the features page.

2. Let it clean

   Cleaning starts on its own, on the phone, with the default settings: 80% intensity, loudness normalization and the 80 Hz low-cut filter.

3. Compare and adjust

   Switch between Original and Cleaned at the same moment, move the intensity slider if the voice sounds processed, and trim the start and the end.

4. Export and share

   Export a file with no watermark (M4A by default, or WAV, MP3 or, on Android 10 or later, Opus), then share it to any app or save it to your phone. Every cleaning stays in History, ready to reopen, adjust and export again.

The same flow cleans the soundtrack of a video: only the sound is processed, never the picture.

Why on-device cleaning matters for personal recordings

An interview with a source, a meeting about a client, a lecture, a message to your family: much of what people record is not meant for anyone else. Noise Remover cleans it on the phone itself, and it works in airplane mode.

• No upload. The app has no server to send audio to, so a recording is never handed to an online service to be cleaned.
• No account. Nothing ties your recordings to a name or an e-mail address.
• Kept on the phone. Your cleanings, originals and exports alike, stay in the app’s storage, excluded from cloud backup, until you delete them.

What does leave the phone (ads in the free version, usage statistics and crash reports) never contains audio. The privacy policy lists each of them.

What every use case has in common: one voice at a time

The engine inside Noise Remover is RNNoise, Xiph.Org’s open-source neural network for speech. It keeps what sounds like a voice and turns down everything else. That is why one voice over fans, hum, traffic, wind or keyboard clicks comes out clean, and why the app cannot pick one voice out of several talking at once.

• Works well: one person speaking, and interviews or meetings where people take turns.
• Works poorly: voices talking over each other, a café full of chatter, background music or TV.
• Not removed: room echo. Record closer to the voice, or in a softer room. See echo or background noise?