Record a video voiceover with timed takes and background music

Prepare the video

Add a browser-playable MP4, MOV or WebM up to 50 MB and ten minutes. Videos above that limit are rejected before file reading or inspection. Open Video Compressor and export a standard H.264 MP4 under the limit, then return here. No media is transferred automatically between the tools. The video preview always shows the complete source aspect ratio.

Record continuously or in intervals

Move the playhead and select Continuous to record until Stop or the video ends. Timed interval uses exact video start/end times and stops at its end. Enable automatic advancement for repeated 5, 10, 15, 30 or 60-second windows. A count-in gives time to get ready. Permission is requested only after Record; canceling the preparation releases any subsequently granted stream. The microphone is closed after every stop, cancellation and unmount.

The recorder captures microphone audio separately from the video. Pre-play recording is trimmed against the video and audio clocks. A take cannot continue through a seek, buffer stall, pause, page backgrounding or playback-rate change, because doing so would lose synchronization. The available recording is saved when feasible. Mobile operating systems and hardware can still introduce latency; use the ±50 ms nudge controls and review an encoded sample.

Edit takes without losing source audio

Select a segment on the voice rail or use the Takes selector. Place in video moves the take, while Source starts/ends trim the original microphone recording. Small edge fades reduce clicks. Source edits are non-destructive; Raw voice downloads the original recording, without timeline trimming or volume changes. Retake records the selected window and only replaces its source after the new recording has been decoded and validated. Overlapping active takes intentionally play together; mute or remove alternatives.

Mix and export

Original sound, voice takes and music have independent levels. Music can loop a selected range from an exact position. Ducking follows the edited voice intervals, with short attack and release ramps; it is timeline-based, not speech recognition. Voice clarity is an 80 Hz high-pass filter. Browser microphone noise suppression, echo cancellation and gain options apply when recording and may vary by device. Monitoring is off by default to avoid speaker feedback; use headphones if enabled.

The live mix schedules source buffers on one Web Audio clock and resynchronizes against the video. A finite preview cache protects memory; muted or deleted takes are excluded. Export releases decoded preview buffers and uses the original compressed sources. Encode a ten-second sample near the playhead to hear the final peak limiter and AAC compression. The full MP4 uses H.264/AAC, the original aspect ratio, no upscaling and a maximum long edge of 1920 pixels. Save all results before leaving; there is no automatic project storage.

Privacy and limits

Video: 50 MB and 10 minutes, up to 40 megapixels per frame. Music: 25 MB / 10 minutes. Voice: up to 32 takes, 25 MB combined and 20 minutes of recorded sources; overlaps mix together. Recording needs HTTPS, microphone permission and MediaRecorder support. The page must stay active; backgrounding, video buffering, seeking, pause, input disconnection and speed changes stop the take to avoid drifting out of sync. Canceling permission does not grant microphone access, and late permission results are stopped. Browser microphone noise suppression, echo cancellation and automatic gain depend on device support. Headphones are recommended for optional mix monitoring; microphone audio is never sent directly to the speakers. Voice files use browser-supported Opus/WebM, Ogg or AAC/MP4 recording. Live preview uses Web Audio with a bounded 48-million-sample cache, and can differ from the encoded peak-limited AAC result. Final MP4 preserves aspect ratio, never upscales, caps the long edge at 1920 pixels and uses 24 or 30 fps. Source metadata is removed. No project persistence is provided: download the result and raw takes before leaving or replacing the video. First use downloads the existing FFmpeg media engine (about 31 MB) from a CDN; video, music and microphone recordings are processed locally, without media uploads. Long exports may exceed the device’s memory or the engine timeout; use a shorter video or lower output size.