# AI audio denoiser: denoise audio and video

Denoise speech with a 48 kHz filter that keeps your real voice intact.

Powered by [Remove noise](https://audo.ai/noise-removal.md) — non-generative 48 kHz spectral mask.

## How it works

1. Upload a lossless WAV or FLAC stem, MP3 voice recording, or 4K video up to 10 GB.
2. Set Denoise amount between Subtle (10%) and Maximum (100%), and toggle Normalize volume as needed.
3. A/B audition Original beside Denoised in the player, then export your stem or untouched video.

## Production workflows built for non-generative denoising

Attenuate broadband hiss, room rumble, and mic floor at 48 kHz without letting a speech synthesizer alter vocal identity.

### Zero voice resynthesis: Avoid synthetic lisps, shifted accents, and swallowed laughs

Generative speech rebuilders such as Adobe Enhance and Descript Studio Sound synthesize a brand-new vocal waveform, which can introduce artificial sibilant lisps, flatten British, Irish, or regional accents, clip genuine laughter, or invent phantom syllables. Because this subtractive AI denoiser only attenuates noisy bins, your exact vocal delivery stays intact — compare models in [best AI noise removal tools](https://audo.ai/guides/best-ai-noise-removal-tools.md) and [Adobe Podcast Enhance](https://audo.ai/guides/adobe-podcast-enhance.md).

### DAW & NLE dialogue stems: Export lossless 48 kHz WAV and FLAC without $99–$399 VSTs

Dialogue editors working in Pro Tools, Logic Pro, Premiere Pro, or DaVinci Resolve can export sample-accurate 48 kHz WAV or FLAC stems straight from the browser without purchasing $99 to $399 per-seat plugins or managing iLok activations — see [Waves Clarity VX and Supertone Clear alternatives](https://audo.ai/guides/waves-clarity-vx-supertone-clear-alternative.md).

### Audiobooks & voice cloning: Keep genuine human formants and a steady noise floor

ACX audiobook chapters, commercial voiceovers, and neural voice-cloning reference datasets require uncolored human vocal formants rather than AI-reconstructed speech. Follow our guide to a [clean voice sample for cloning](https://audo.ai/guides/clean-voice-sample-for-cloning.md), then verify integrated loudness and true peak in [LUFS meter](https://audo.ai/lufs-meter.md).

### 4K video stream-copy: Denoise MP4, MOV, and MKV soundtracks up to 10 GB directly

Skip bouncing separate audio stems out of your timeline: drop MP4, MOV, or MKV masters up to 10 GB straight in and export H.264, HEVC, or Apple ProRes video streams copied bit-for-bit — or open [Remove noise from video](https://audo.ai/remove-noise-from-video.md) for camera-first shoots.

## 48 kHz spectral gain masking vs. generative voice resynthesis

Audo's online audio denoiser — built on [Remove noise](https://audo.ai/noise-removal.md) — processes uncompressed and lossy audio (WAV, MP3, M4A, AAC, FLAC, OGG, Opus, WebM, AIFF) plus video containers (MP4, MOV, MKV, WebM, AVI) up to 10 GB as a guest (multipart over 100 MB, and up to 10 hours signed in), with the first minute of each file free, 3 times a day, with no sign-up. Measured audio turnaround averages about 7 seconds for 1 minute, about 33 seconds for 20 minutes, and under 2 minutes (about 1 minute 41 seconds) for a 1-hour master.

Under the hood, a non-generative 48 kHz spectral mask operates strictly by subtraction: for every short-time frequency-time bin, the network predicts a bounded attenuation gain between 0 and 1, multiplying down noisy magnitudes while leaving the recording's original phase untouched and never synthesizing new harmonics above 1. Vocoder-based speech enhancers work the opposite way — decoding phonetics and generating a synthetic studio voice from scratch, which explains why heavy resynthesis can smooth away vocal rasp, alter regional vowels, or hallucinate consonants in noisy passages.

Rather than leaving Denoise amount pinned at Maximum (100%) on every project, try dialing the 10–100% wet/dry blend to 75–85% on quiet podcasts and audiobooks so breaths and laughs sit over a gentle, natural noise floor (uncheck Normalize volume to preserve raw timeline gain, or audit levels afterward in [LUFS meter](https://audo.ai/lufs-meter.md)). When a recording suffers from hollow bare-wall bounce or a thin, muffled capsule, switch to [Audio enhancer](https://audo.ai/audio-enhancer.md) or [Remove echo](https://audo.ai/remove-echo.md); when a single transient intrusion like a dog bark or siren ruins a take, isolate it with [Sound remover](https://audo.ai/sound-remover.md).

How five denoising approaches treat vocal timbre, phase, and picture

| Denoising approach | Effect on the voice & file | Best suited for |
|---|---|---|
| Non-generative 48 kHz spectral mask (this page) | Applies a [0, 1] gain mask per bin; preserves original phase, accents, and laughs | Dialogue stems, podcasts, audiobooks, and voice-cloning reference audio |
| Generative speech resynthesis | Reconstructs 48 kHz vocal harmonics from scratch; removes echo and mic thinness | Distant, muffled, echoed, or clipped recordings in [Audio enhancer](https://audo.ai/audio-enhancer.md) |
| Static noise-print FFT gate | Subtracts one frozen noise profile; leaves musical chirp when background shifts | Strictly constant analog tape hiss with a clean silent sample |
| Prompt-guided single-sound separation | Targets one described sound while leaving surrounding scene ambience intact | Removing sirens, barks, or bells via [Sound remover](https://audo.ai/sound-remover.md) |
| Visual pixel-grain video denoisers | Smooths high-ISO camera sensor grain on image frames; never touches audio | Low-light footage cleanup inside an NLE or color suite |

Tool: Remove noise (`remove_noise`). The same tool's other pages: https://audo.ai/noise-removal.md, https://audo.ai/remove-noise-from-video.md, https://audo.ai/voice-isolator.md, https://audo.ai/remove-static.md, https://audo.ai/remove-wind-noise.md, https://audo.ai/remove-fan-noise.md. Price: 5 credits a minute — $0.05/min ($3.00/hr), pay-as-you-go with no monthly subscription — of input for whole files, charged only when a job succeeds.

## Free API

```bash
curl -F file=@dialogue-stem.wav -F noise_reduction=85 https://audo.ai/api/try/remove-noise -D response.headers -o response.body
```

Read response.headers and response.body before naming media. Handle 202, 200 JSON and errors with the complete response client at https://audo.ai/docs/api?utm_source=markdown&utm_medium=agent#responses.

Free for the first minute of each file, 3 times a day per person, with no key or account. Uploads up to 10 GB, 2 at once from one address. The response is the result itself; a job that runs longer than about 30 seconds returns 202 with a link to check on it. With an API key (`-H "Authorization: Bearer $AUDO_API_KEY"`), omitting free runs the full job with credits if the balance covers it, otherwise the free allowance if quota remains. Set free=true to prohibit spending; for paid runs use free=false with an approved positive max_credits. A free result cut short comes with finishUrl (and the Audo-Finish-Url header when the response is the file itself): give it to the person, who pays there (from $5, no sign-up), and the whole file runs; keep checking the status link, whose finishedBy names the whole-file job and its own status link. Details: https://audo.ai/docs/api.md

## Options

Form fields with the same names as the MCP tool's inputs:

- `noise_reduction`: How much of the noise to turn down, as a percentage. 100 turns down all it can; lower values keep some of the background. Default 100.
- `auto_volume`: Sets the whole file to -16 LUFS with one gain, and limits peaks to -1 dBTP. It doesn't balance loud and quiet speakers. Default true.
- `output_format`: The result's format: same (as the input; a video stays a video), wav, mp3, m4a, or flac. Default same.

## Languages

Any language: it works on the sound, not the words.

## From an assistant (MCP)

Add https://audo.ai/mcp and sign in (setup for each assistant: https://audo.ai/setup.md). The tool is `remove_noise`:

> Turns down background noise, hiss, hum, and room noise behind speech, keeping the speaker's real voice: it filters the recording and adds no sound of its own. Then sets the file to -16 LUFS. Made for voices: music and other sounds are turned down too. Works on video. Use it when the voice is fine but there's noise behind it. Not for echo, phone, or damaged audio (enhance_voice). It costs 5 credits a minute, charged only if the job succeeds. Returns the result with download links, or job_id for get_job if it takes longer.

Example: "Use Audo to denoise dialogue-stem.wav at 85% strength and keep the original vocal tone."

## Questions

**Is there a free denoiser to denoise audio free and download the file?** Yes. Powered by Audo's Remove noise, this free audio denoiser and MP3 denoiser lets you denoise audio free on the first minute of each audio or video file, 3 times a day, with no sign-up — returning a real downloadable file, unlike cleanaudio.io, noise-reducer.com, and lalal.ai, which block free downloads behind a paywall. Full-length files cost $0.05 a minute ($3.00 an hour), pay-as-you-go from $5 with a non-expiring balance and no subscription — [see pricing](https://audo.ai/pricing.md).

**What does it mean to denoise audio without resynthesizing the voice?** When you denoise audio on Audo, a 48 kHz spectral mask scales each time-frequency bin by a factor capped between 0 and 1 while keeping the original phase intact. Because zero artificial harmonics are generated, the speaker's true vocal formants, regional accent, breath cadence, and laughter never shift into an AI-generated clone.

**Why keep Denoise amount around 75–85% for podcasts, laughs, and audiobooks?** At 100%, heavy background attenuation can strip every trace of room tone between phrases, making sudden breaths or laughs hang over dead digital silence. Setting Audo's 10–100% Denoise amount slider around 75–85% blends a controlled, natural noise floor back under speech. Normalize volume is switched on by default to set output to −16 LUFS with a −1 dBTP true-peak ceiling — uncheck it to preserve original DAW gain, or inspect exact integrated loudness in [LUFS meter](https://audo.ai/lufs-meter.md).

**How does Audo compare with browser denoisers and VST denoiser plugins?** Many browser denoisers cap uploads at 2 GB or 60 minutes and refuse free file downloads, while desktop plugins cost $99 to $399 upfront plus per-machine iLok licensing and local GPU load. Audo accepts lossless WAV, FLAC, and 4K video up to 10 GB in the browser (up to 10 hours signed in) with downloadable free runs and pay-as-you-go billing — compare tools in [Waves Clarity VX and Supertone Clear alternatives](https://audo.ai/guides/waves-clarity-vx-supertone-clear-alternative.md).

**Does this denoise grainy or pixelated low-light video footage?** No. Audo denoises the audio soundtrack only — it does not smooth visual ISO sensor grain, chroma blotches, or compression macroblocking on video pixels. When you upload MP4, MOV, MKV, WebM, or AVI footage with Output format set to Same as the original, Audo cleans the soundtrack and copies your video frames across bit-for-bit untouched.

**What if my recording also has room echo or muffled speech?** Because non-generative spectral masking preserves how the voice hit the capsule, delayed wall reflections and missing upper vocal frequencies remain in the signal. To dry out bare-room reverb or reconstruct crisp highs on a muffled microphone, run Audo's [Audio enhancer](https://audo.ai/audio-enhancer.md) or [Remove echo](https://audo.ai/remove-echo.md) instead.

**Can I automate batch audio denoising via REST API, curl, or MCP?** Yes. Automate single-call or multi-file batch pipelines with Denoise amount and Output format parameters via Audo's [noise removal API](https://audo.ai/docs/noise-removal-api.md), or connect Audo's MCP server so ChatGPT, Claude, Claude Code, and Cursor can denoise recordings directly from chat. [Add Audo to your assistant](https://audo.ai/setup.md) once to enable it.

## More

- Other clean-ups: [Audio enhancer](https://audo.ai/audio-enhancer.md), [Remove noise](https://audo.ai/noise-removal.md), [Remove echo](https://audo.ai/remove-echo.md), [Remove music from video](https://audo.ai/remove-music-from-video.md), [Sound remover](https://audo.ai/sound-remover.md), [Remove static](https://audo.ai/remove-static.md), [Fix muffled audio](https://audo.ai/fix-muffled-audio.md), [Remove wind noise](https://audo.ai/remove-wind-noise.md), [Fix distorted audio](https://audo.ai/fix-distorted-audio.md), [Podcast audio enhancer](https://audo.ai/podcast-audio-enhancer.md), [Remove fan noise](https://audo.ai/remove-fan-noise.md), [Mic enhancer](https://audo.ai/mic-enhancer.md), [Voice isolator](https://audo.ai/voice-isolator.md), [Audio cleaner](https://audo.ai/audio-cleaner.md), [Voice cleaner](https://audo.ai/voice-cleaner.md)
- Related guides: [Remove background noise in Premiere Pro](https://audo.ai/guides/remove-background-noise-premiere-pro.md), [Remove background noise and echo in Audacity](https://audo.ai/guides/remove-background-noise-audacity.md), [Best AI noise removal tools](https://audo.ai/guides/best-ai-noise-removal-tools.md), [iZotope RX Voice De-noise & plugin alternatives](https://audo.ai/guides/izotope-rx-voice-de-noise-alternative.md), [Adobe Audition noise reduction & DeNoise](https://audo.ai/guides/adobe-audition-noise-reduction.md), [ACX noise floor (−60 dB) & audiobook cleanup](https://audo.ai/guides/acx-noise-floor-audiobook-cleanup.md)
- [All tools](https://audo.ai/all-tools.md)
- [Prices and free limits](https://audo.ai/pricing.md)
- [Set up your assistant](https://audo.ai/setup.md)
- Every tool: https://audo.ai/llms.txt
- OpenAPI: https://audo.ai/api/openapi.json
- This page for people: https://audo.ai/audio-denoiser?utm_source=markdown&utm_medium=agent
