Remove Background Noise from a Video
Pick a video and a strength — background hiss, hum and noise are reduced in your browser while the picture stays untouched. Nothing is uploaded.
Clean up a noisy video's audio
Pick a video, choose how hard to reduce the background noise in its audio, and get the same clip back — clearer. The picture is stream-copied (not re-encoded), so it stays byte-for-byte identical and processing is fast; only the audio track is re-encoded. Everything runs in your browser; nothing is uploaded.
Choosing a strength
The Strength slider (1–100) controls how aggressively noise is removed:
- Low (10–20) — a gentle cleanup that shaves off steady hiss while keeping voices natural. This is the safe starting point.
- Medium (30–50) — a stronger reduction for noticeable air-conditioner hum, fan noise, or tape hiss.
- High (60–100) — maximum reduction. Effective on heavy noise, but push too far and quiet passages can start to sound hollow or "underwater".
Start low and raise it only until the noise is gone — over-processing is the usual mistake.
Worked example: a talking-head clip recorded near a noisy laptop fan. Load
interview.mp4, set Strength to 20, leave Denoiser on FFT
(afftdn), and you get interview-denoised.mp4 — the same video with the fan
hiss pulled down and the voice intact.
Denoiser and hum removal
- Denoiser — FFT (afftdn) is the fast, general-purpose default and works well for steady hiss/hum. Non-local means (anlmdn) is slower but can hold onto fine detail on broadband noise; try it if afftdn sounds too smeared.
- Also remove low hum/rumble — adds a high-pass filter at 80 Hz to cut low-frequency mains hum, HVAC rumble and handling thumps. Leave it off if your audio has wanted bass (music, deep voices).
Notes and limits
- The video stream is copied losslessly; the output keeps the same container (mp4 → mp4, webm → webm). WebM audio is re-encoded to Opus, everything else to AAC.
- This is a classic ffmpeg denoiser, not an AI "studio voice" model — it reduces steady background noise rather than isolating a single speaker.
- One setting is applied to the whole clip. To also raise the level afterward, run the result through the Change a Video's Audio Volume tool.
- Input and output are each capped at 25 MB (the file is processed in your browser's memory).
FAQ
Is my video uploaded to a server?
No — ffmpeg runs inside your browser tab, so the file never leaves your device.
Will denoising hurt the video quality?
No. Only the audio is changed; the picture is stream-copied without re-encoding, so the video quality is identical to the original.
What strength should I use?
Start around 12–20 and raise it only until the noise is gone. High values (60+) remove more noise but can make quiet parts sound hollow or robotic, so increase gradually and stop when it sounds clean.
What's the difference between the FFT and non-local means denoisers?
FFT (afftdn) works in the frequency domain and is fast — a good default for steady hiss and hum. Non-local means (anlmdn) compares similar chunks of the waveform and can preserve fine detail better on broadband noise, at the cost of speed. If one sounds too smeared, try the other.
Can it remove a hum or buzz?
Turn on Also remove low hum/rumble to add an 80 Hz high-pass that cuts mains hum, HVAC rumble and handling thumps. Combine it with a moderate strength for the rest of the noise.
Which video formats can I use, and how big can the file be?
Anything ffmpeg can read — mp4, mov, mkv and webm are the common cases. The
output keeps the input's container and is named after the original with a
-denoised suffix (e.g. clip.mp4 → clip-denoised.mp4). The input and output
are each capped at 25 MB.
Developer & Automation Access
Run it from the terminal
Same engine as this page, headless — via the gizza CLI:
gizza tool video-audio-denoise 'url=https://example.com/input' 'strength=12' 'method=afftdn' 'remove_hum=true'New to the CLI? Get gizza →
Open it by URL
Pre-fill and auto-run this tool with query parameters — the names match the API/CLI:
https://gizza.ai/tools/video-audio-denoise/?url=https://example.com/input&strength=12&method=afftdn&remove_hum=trueMachine-readable descriptor: tool.json — title + parameters JSON Schema for agents.