Free Auto Ducking — Voice Over Music, No Upload
Quick answer: Automatically lower music when you speak — free online auto ducking for podcasts and videos. Processed privately in your browser, no upload needed.
Auto ducking is the process of automatically lowering the volume of background music whenever a voice starts speaking, then smoothly bringing it back up when the voice stops. Audio Forge Pro does it entirely in the browser the way Auphonic and Premiere's Auto-Duck do: it pre-analyses the voice with a hybrid speech detector (neural VAD + RMS envelope), classifies the timeline into speech / music / silence, then writes a look-ahead gain curve with an adaptive release — fast and full in long pauses, held through short ones. Drop your voice on the top lane, your music on the bottom lane, pick a content type, drag any keyframe, and export. No upload, no watermark.
Audio Forge Pro is a free browser-based auto ducking studio built by M Zeshan (ShanX). It pre-analyses your voice with a hybrid Silero neural VAD + spectral DSP detector, writes a look-ahead keyframe ducking curve with adaptive release, and supports both broadband ducking and Smart Vocal Carve entirely in your browser. Export the full mix or a ducked music stem as MP3 320 kbps or lossless WAV with a true-peak ceiling around -1 dBTP.
How to Auto Duck Voice Over Music (3 Steps)
- Drop voice on the top lane, music on the bottom: Two dashed lanes, both decoded locally in your browser. Nothing is uploaded — voice and music stay on this device.
- Pick a content type, preview live, drag the curve: Eight presets (vlog, narration, podcast, long-gap commentary, ASMR, gaming, tutorial, roundtable) set duck depth, attack, adaptive release and look-ahead. The duck curve is drawn on the waveform — preview it without rendering and drag any keyframe.
- Press Duck & mix and download MP3 or WAV: The whole timeline is rendered offline in your browser with a true-peak ceiling of about -1 dBTP so the result never clips. Download starts straight to your Downloads folder.
What Auto Ducking Supports
| File | Notes |
|---|---|
| Voice — MP3 | Decoded locally in the browser · typical for podcast exports |
| Voice — WAV | Lossless voice master · sample-accurate start |
| Voice — M4A / AAC | iPhone Voice Memos, WhatsApp audio, mobile captures |
| Music — MP3 | Most royalty-free music beds · 320 / 192 / 128 kbps |
| Music — WAV | Lossless music bed · recommended when available |
| Output — MP3 320 kbps | Default · transparent for voice and music |
| Output — WAV | Lossless · for further editing in a DAW |
Who Uses Auto Ducking
- Shorts voiceover over a music bed — 15–60 second vertical videos.
- YouTube explainer with a quiet theme tune — 5–20 minute videos.
- Podcast intro / outro music — 30-second bed over a 5-minute cold open.
- Audiobook narration with subtle ambient music — long-form voice.
- Tutorial screen-record voice — music in the intro, voice in the body.
- Gaming commentary with BGM — menu music ducks during play-by-play.
Honest about what a browser duck can do
- Pre-analysis ducking: hybrid neural VAD + RMS speech detection, speech / music / silence segments, look-ahead gain keyframes with adaptive release.
- Eight content-type presets (vlog, narration, podcast, long-gap commentary, ASMR, gaming, tutorial, roundtable) — sliders and draggable keyframes for fine-tuning.
- Loop the music, trim it to the voice, or fit to the shorter track — your call.
- Negative music pre-roll so the bed is already moving when the voice starts.
- True-peak ceiling of about -1 dBTP so the rendered file never clips.
- No upload, no watermark, no account, no daily credits.
Auto Ducking — Frequently Asked Questions
What is auto ducking?
Auto ducking is the process of automatically lowering the volume of background music whenever a voice starts speaking, then bringing the music back up when the voice stops. Audio Forge Pro analyses the whole voice file first (speech / music / silence), so the music starts ducking 100–250 ms before each phrase, stays ducked through short pauses, and recovers fast and fully in long gaps — no late reaction, no buried music.
How do I duck background music under a voiceover online?
Drop your voice file on the top lane of the auto ducking tool, drop your music on the bottom lane, pick a content type (or accept the suggested one), preview live, and click Duck & mix. The mix is rendered offline in your browser and the MP3 or WAV downloads straight from your device. No upload, no account, no watermark.
Is this a sidechain compressor?
It does what a sidechain compressor does, but better for speech: instead of reacting to level alone it uses a voice-activity detector (Silero neural VAD with a spectral fallback) fused with a 20 ms RMS envelope, so loud music beats or SFX in the voice track do not trigger a duck and a whisper still does. Duck depth, music bed level, attack, release, look-ahead, hold and gap bridging are all adjustable.
Do you upload my voice or music to a server?
No. Decoding, envelope analysis, and the ducked mix are all rendered entirely in your browser. Files never leave the device. The only network activity is downloading the static page and the MP3 you choose to save.
Will the mix click or pop when the voice starts and stops?
No. Every gain move is an exponential ease-out (the same law as Web Audio's setTargetAtTime): quick while it is far from the target, gentle as it lands. Pauses shorter than the bridge time stay ducked so fast talking never pumps, and long pauses recover 90 % of the duck within the release time (250–600 ms depending on the preset).
What happens if the music is shorter than the voice?
By default the music loops with a 5 ms crossfade at the boundary, so a 30-second music bed can carry a 60-second voiceover. The Studio UI also lets you trim the music to the voice length or fit the output to the shorter of the two. Loop works best for Shorts, trim is the safer default for podcasts.
What if the music is longer than the voice?
The default is to trim the music where the voice ends. If the music has a strong outro you want to keep, switch to “Fit to the shorter track” — both inputs stop at the shorter duration, so nothing trails off awkwardly.
Can I start the music before the voice comes in?
Yes. The music start offset slider (under Advanced) lets you set a pre-roll so the music is already in motion when the voice begins. Independently, every preset has 100–250 ms of look-ahead so the duck is already in place when the first word lands.
Does it work for YouTube Shorts and TikTok?
Yes. Use the Fast talking / Vlog preset (20 ms attack, 350 ms release, 10 dB duck) or Energetic / Gaming (15 ms, 300 ms, 15 dB) for hype edits. The music is at full bed level before the first word, gets out of the way the moment you speak, and eases back in at the end. For 15–60 second vertical videos the same browser tab on a phone works fine; for longer content render on desktop.
Does it work for 20–90 minute podcasts and YouTube videos?
Yes, on desktop. The tool accepts mixes up to 90 minutes; the Podcast / Interview preset is tuned for long-form dialogue with a 10 dB duck, 40 ms attack and 500 ms release. Very long mixes take a minute or two to analyse and render because the whole file is built in memory — keep the tab open until the download appears.
Can I use this with my own music and voice, or do I need royalty-free files?
Use whatever files you own or are licensed to use. The tool does not add music or generate a voice; it just mixes two files you already have. You stay responsible for the rights to both the voice and the music in the final output.
How does this compare to a desktop DAW sidechain?
A desktop DAW (Adobe Audition, Reaper, Logic, Pro Tools) gives you more surgical control — multiband sidechain, parallel compression, sidechain EQ — and ships in a paid app. Audio Forge Pro focuses on the one workflow creators actually need (drop voice + music, render duck, download) and runs it locally for free. If you already have a DAW, use it; if you do not, this is a quick, honest alternative.
AI Silence Remover | -14 LUFS Normalizer | Audio Trimmer | Audio Joiner | Extract Audio from Video | Full Audio Studio
All Audio Forge Pro Tools
- AI Silence Remover — Remove dead air automatically
- -14 LUFS Normalizer — YouTube & Spotify loudness standard
- Stereo to Mono Fixer — Fix single-ear left audio & downmix
- Audio to Video — Waveform audiogram for Shorts & clips
- Extract Audio from Video — MP4 to MP3 — no upload, local encode
- Audio Joiner — Merge MP3 files on a timeline
- Audio Trimmer — Cut MP3 — split, delete, reorder
- Auto Ducking — Voice over music — music ducks when you speak
- Remove BG Music — AI voice/music split — free, unlimited
- Add Echo to Audio — Slapback, tape, ping-pong & karaoke echo presets
- Add Reverb to Audio — 10 convolution spaces — room to cathedral
- Slowed and Reverb — Slowed edits with hall reverb & warm bass
- Sped Up Song Maker — 1.15x TikTok speed, or keep the pitch
- Nightcore Maker — 1.3x speed + pitch — the classic sound
- 8D Audio Maker — Rotating pan + room — headphones on
- Audio Speed Changer — Whole song or one section — speed and pitch
- Voice Echo Effect — Speech-tuned echo and plate for vocals
- Full Audio Studio — All tools in one workspace