What a voice-over picks up, and what removes it
A narration is one voice, close to the microphone, in a quiet room: the situation Noise Remover handles most cleanly. What is left to remove is the room itself, and part of it is not noise at all:
| Sound | Where it comes from | What removes it |
|---|---|---|
| Hiss | A microphone with its gain turned up, an older recorder or audio interface | The cleaning, which handles it well |
| Room tone | Computer and air-conditioning fans, a fridge, the building’s ventilation | The cleaning (see fan noise) |
| Hum | Mains wiring, a charger, a lamp dimmer near the cable | The cleaning |
| Low rumble | Desk thumps, traffic through the floor, the bass boost of a voice very close to the microphone | The 80 Hz low-cut filter, then the cleaning |
| Echo | Bare walls, a hard floor, an empty room | Nothing in the app: treat the room before you record |
In our lab test at full intensity, hiss dropped by 23.2 to 42.9 dB in the speaker’s pauses, while the voice’s own level changed by only 0.2 to 0.6 dB (how we measured).
| Noise | Noise removed during pauses, in dB | Change in voice level |
|---|---|---|
| Hiss, white noise | 23 to 43 dB | −0.2 to −0.6 dB |
| Electrical hum | 24 to 41 dB | −0.3 to −1.6 dB |
| Broadband noise | 26 to 39 dB | −0.3 to −0.9 dB |
| Low rumble | 25 to 34 dB | −0.3 to −1.0 dB |
| Fan, air conditioning | 24 to 25 dB | −0.3 to −1.5 dB |
How to record a take the cleaning can work with
- Deal with the echo first. The cleaning does not remove it. A wardrobe full of clothes, a duvet hung behind the phone or a sofa corner absorbs reflections far better than a bare desk. See echo or background noise?
- Keep one distance. About 15 cm (6 in) from the mouth, slightly to the side to avoid breath pops, and the same for every take, so your timbre does not shift between paragraphs.
- Leave headroom. Speak at your normal level rather than pushing the microphone to its limit: loudness normalization brings the take up to −16 LUFS afterwards.
- Skip the noise sample. Profile-based noise reduction asks for a few seconds of room tone before you speak; Noise Remover does not, because its model follows the noise as it goes. A short pause still makes the start easier to trim.
For a closer, fuller sound, on Android 10 or later, plug in a wired or USB microphone and choose it in the recorder before you start: a dedicated microphone near your mouth hears far less of the room than the phone does at arm’s length.
Drop a marker each time you fluff a line, and the retakes are easy to find when you trim or edit.
More recording advice, from where to put the phone to when to record, is in how to record clean audio with your phone.
How do I keep my voice natural after noise reduction?
Narration is often heard on headphones, where over-processing is easy to spot: a thin, lisping or watery voice. Start from the default 80%, which caps the reduction near 29 dB, and use the Original/Cleaned switch on the passages that give processing away first: soft endings, breaths, “s” sounds. If the voice loses body, lower the intensity; if room tone is still audible in the pauses, raise it. More in why noise reduction can sound robotic.
Hear the difference
A voice over the noise of a fan, cleaned by the app’s own engine. Press play, then switch between Original and Cleaned.
The 80 Hz low-cut filter removes rumble below the voice. With a very deep voice, compare with the filter off; switching it re-runs the cleaning, which takes a moment on a long take.
Finish it for YouTube, a course or a client
Loudness normalization (Ready-to-publish level in the app) brings the narration to −16 LUFS with a −1 dBTP ceiling, which keeps a series of videos or the lessons of a course at one level from one recording session to the next. Trim the breath before the first word and the silence after the last, then export an M4A (AAC at 128 kb/s, 48 kHz, no watermark) for a YouTube or explainer edit, a course platform or a client review.
Many voice-over clients ask for WAV deliveries. Export WAV when they do: it is uncompressed, and it is also the file to hand an editor who will encode the final video.
Voice-over is normally delivered in mono, and a recording made in the app is mono: it stays mono. A stereo file you import, from an audio interface for instance, stays stereo unless you choose Mono in the export options, which folds it to the middle. Keep mono unless your template or your client asks for a two-channel file.
On the free plan, the first 3 minutes of each file are cleaned, 3 files a day: enough for a short explainer or a few sections of a course. A longer take needs the optional video, which cleans one file up to 15 minutes, or Pro, for any length and no daily limit.
Clean your next take on your phone.
Free · Android 8.0 or later, 64-bit
What stays in a voice-over after cleaning
- Room echo. It stays, and a cleaned voice in a bare room still sounds far away.
- Music already under the voice. It is pushed down and altered. Record the voice dry and add the music in the edit.
- A recording that was already clean. With no noise to remove, cleaning can colour the sound slightly. The Original/Cleaned switch tells you which one to keep.
- Noise as loud as the voice. The noise goes, but the voice comes out thinner. A quieter hour or a closer microphone helps more than a higher intensity.