Click or drag videos, audio and images here
MP4, MOV, MKV, WebM, MP3, M4A, WAV, OGG, FLAC, PNG, JPEG, WebP — nothing leaves your browser
Everything here applies to every clip that has not been given settings of its own. Open a clip to change it alone; a clip with its own settings says so on its row, and can be handed back at any time.
How the Video Editor works
Free Online Video Editor — Cut Silence, Zoom, Caption and Export
Arrange video, audio, images and text cards into one timeline, cut the silence out of every clip, push in on the picture after the longest pauses, change the speed, replace the sound, add a caption along the bottom and export the result as MP4, WebM, MKV or MOV — or as audio alone in M4A, MP3, OGG or WAV.
Everything runs inside your browser using WebCodecs. Your files are never uploaded, and nothing is stored anywhere but on your own computer.
Instructions
How the timeline works
Drop your files onto the upload area and they appear as a list. The list is the timeline: clips play in the order shown, back to back, with no gaps. Drag a clip by the handle on the left, or use the up and down arrows on its row. Between any two rows there is a plus: it asks whether you want a file or a text card and drops it at exactly that position, which is usually where a title belongs. "Add text card" at the top appends one at the end instead, and any clip can be duplicated with the copy button — useful when you want the same footage twice at two different speeds.
Watching it before you export
"Preview" plays the whole timeline as it stands, without writing anything. It is not an approximation: every frame goes through the same function the encoder uses, so the cuts, the speed, the fades, the zooms, the captions and any replaced or continued sound are exactly what will end up in the file. The one thing it cannot show is volume levelling, which is measured from an analysis and applied sample by sample while exporting.
Under the picture is the clip currently on screen, with its settings open — so you can watch a zoom land, change how far it pushes in, and see the difference on the next pass without leaving the preview. Under that is a strip of every clip in order, each showing its first frame. It scrolls sideways when there are more clips than fit, follows the playhead as the video runs, and any card can be dragged to a new position. Click a card to open that clip's full editor, or use the small play button on it to jump the preview there.
Settings for everything, then for one clip
The panel at the top holds every setting there is, and it applies to every clip that has not been given settings of its own. Open a clip with "Edit" and change anything inside it, and that clip quietly takes a copy of the project's settings and keeps it — its row then says own settings. Tick "Follow the project settings" inside the clip to hand it back, and it returns to whatever the project says, including anything you changed while it was away.
There is no per-field inheritance and no third state on purpose. "Which of these eleven numbers is still following the project?" is a question nobody should have to answer in the middle of editing.
Cutting the silence, and cutting by hand
Tick "Remove the silent stretches" and every clip that follows the project has its pauses found and removed. The detection settings behind it are the ones from the Free Silence Cutter: how quiet counts as quiet, how long a pause has to last before it is worth removing, and how much room is kept in front of and after each word. Press "Listen to this clip" inside a clip to see the result on a waveform before exporting — and if you do not, the clips that need listening to are listened to automatically when you press Export.
On the waveform you can drag across any stretch to cut it by hand, whether or not it is silent, which is how you remove a false start or a section you simply do not want. Click a marked stretch to take it back. A stretch you drew disappears when you click it; a pause the detector found is only switched off, so it stays off the next time the clip is listened to.
Zooming in after a long pause
When a pause is removed, the two halves it joined were filmed with the same framing, so the join shows up as a jump. Tick Automatic zoom after long pauses and the picture pushes in slightly on whatever follows, which hides the jump and gives the edit a rhythm.
Only pauses that stand out are used: the tool takes the average length of everything it removed from that clip and zooms after the ones at least 30% longer than average, so an ordinary gap between two sentences is left alone. The gear beside the checkbox opens the settings — how far above average a pause has to be, how far the picture pushes in (10% to 20% by default), whether that amount is picked at random inside the range or in proportion to how long the pause was, how long the push-in takes and how long it is held. You can also switch the rule to every so many cuts, which zooms by a fixed amount on a count rather than on the content — that is the one to use on a densely cut stretch where every pause is much the same length, so none of them stands out. The random choice is seeded from the clip itself, so the same footage always gets the same zooms and a second export never differs from the first. Nothing is stretched: the picture is enlarged and the edges are cropped.
Speed
Every clip can play from half speed to ten times, in half steps. Speed is applied after the cuts, so a clip whose pauses were removed and which is then run at double speed lasts half of what survived. The sound is resampled along with the picture, which means its pitch rises as it speeds up and falls as it slows down, exactly as a tape does — nothing here tries to hold the pitch steady.
Replacing the sound
Inside a clip, set Sound to "Replace it with a file" and choose an audio file — or a video you only want the audio from. The clip's own soundtrack is discarded and the file plays in its place. The clip's length is what decides: sound longer than the clip is cut at its end, and sound shorter than it is followed by real silence, so nothing after it drifts out of sync. "Silence it" removes the sound without replacing it, and does so by writing real silence rather than leaving a hole, which is what keeps every later clip in place.
A clip that brought no sound of its own — footage shot with the microphone off, a silent screen capture, a still photograph — falls back to the project's default sound rather than to silence. If a supplied track is already running when that clip starts, the track simply keeps running through it, which is better than dropping out for one shot and starting the same file again from the top on the next. Silencing a clip on purpose is untouched by any of this: it stays silent. Text cards are too — a title never had sound to lose, so music arriving under one is a decision you make rather than inherit.
Evening out the volume
Clips recorded on different days sit at different levels, and inside a single clip a speaker drifts — leaning in, turning away, dropping their voice at the end of a sentence. Even out the volume across the project measures how loud the speech actually is in each clip and applies the gain that brings it to a stated target, so every clip arrives at the same level. Behind the gear you can also ask it to follow the drift inside each clip, with a gain that moves slowly enough not to be heard as pumping.
It will not lift silence. Room tone, breathing and microphone hiss all sit below the noise floor, and boosting them to speech level is the single most recognisable way to make an edit sound processed — so anything under the floor keeps the gain of the speech around it instead. A limiter rounds off anything that would overshoot rather than letting it square off into distortion.
Captions and text cards
Each clip can carry a caption along its bottom edge. It is always white, with a black outline and a soft shadow, because the one job a caption has is to stay readable over footage of any colour — you choose the size, how far it sits from the bottom, and whether it fades in and out on its own clock, which can differ from the clip's.
A text card is different: it is a clip in its own right, full screen, with its own background colour or photograph, and it takes the animations from the Text Video Maker — a typewriter, rising letters, words arriving out of focus, or a wipe reveal with a slow zoom. Use a caption to label footage, and a text card to introduce a section.
Fades
Any clip can ease in from black and silence and out again. The picture and the sound ramp together, and the audio ramp is applied sample by sample rather than packet by packet, because holding one volume across a twenty-millisecond packet turns a smooth fade into a staircase you can hear as clicks. A clip too short for two full fades splits itself in half so the two ramps meet instead of overlapping, and no fade ever spills into the clip beside it.
Exporting
"Export video" writes a single file in the format you chose — MP4 and MOV for anything that has to play everywhere, WebM for smaller files at the same quality, MKV for archiving. "Export audio only" writes the same edit without a picture, in M4A, MP3, OGG or WAV, which is how you turn an edited interview into a podcast episode. If your browser supports it you will be asked where to save before the work starts, and the file is then written straight to disk as it is produced rather than being held in memory — which is what lets a long project export in a tab that has only a few hundred megabytes to spare.
Subtitles and transcripts
"Export transcript" listens to the timeline and writes the words out as SubRip (.srt), SubViewer (.sbv), WebVTT (.vtt) or as plain text with or without timings. The timings are those of the finished video, not of the material behind it: a pause the edit removed is gone from the transcript too, a clip playing at double speed has its words where they actually fall, and the file can be uploaded beside the export without being nudged. Each clip's own settings dialog has the same button for that clip alone, timed from its beginning.
Both offer to remove the background noise before the model listens. That cleaning is only ever applied on the way to the recogniser — your clips and your exported video are untouched by it — and it is worth ticking for anything recorded outside a quiet room. The speech model is fetched once, on this machine, and no sound leaves it.
What happens to your files
Nothing is uploaded. Every file is read from your disk by the browser itself, decoded with WebCodecs, and written back out locally. There is no account, no queue and no server involved, which also means the work is limited by your own machine — a long project on a phone will be slower than the same project on a desktop.