Video editor

Why the preview is instant

Every other tool on this site works the same way: set the options, run FFmpeg once, get a file. An editor cannot work like that, because you need to see the change while you are making it.

So this runs two engines over one edit. The preview draws your video into a canvas every frame, with the color work done by the browser's own graphics pipeline and the reframing done by drawing the frame twice, blurred behind and sharp in front. That is fast enough to play in real time. When you press export, the same settings are turned into an FFmpeg command and rendered properly.

The consequence worth knowing: the preview is an honest approximation, not the finished file. Color and framing match closely. The exported render is the real thing, at full resolution, with proper encoding.

Four tabs, one job each

There is a lot in this editor now, and a wall of controls is what makes editing software frightening. So the picture and the timeline are always on screen, and everything else sits behind one of four tabs, each of which says in a sentence what it is for: Cut for chopping the video up, Add things for text, pictures, another video, music and captions, Sound for how loud things are and what they sound like, and Finish for saving your work or making the video.

Nothing is destructive and nothing is final. Undo and Redo sit next to Play, and Ctrl+Z and Ctrl+Y do what you expect, eighty changes deep. Every cut, layer, keyframe, effect and speed change goes through it.

Cutting out the pauses

Cut out the quiet parts listens through the whole recording and keeps only the stretches with sound in them, laid end to end. That is the jump cut every talking-head video is made of, in one press. You choose how quiet counts as a pause and how long a pause has to be before it is cut, so it does not chop between words, and there is a small pad either side so nothing is clipped.

A shelf for things you use often

Under Add things is a shelf. Put a folder of sound effects, a set of overlays or a font on it and they stay there for the rest of the edit, one click from the timeline. Drop files straight onto the box or pick them; they are sorted into sounds, pictures, videos and fonts. Using one puts it at the playhead and leaves it on the shelf, so the same whoosh can go in twenty times without finding the file again.

Doing something to just one bit

Drag across the ruler above the timeline, or right-drag anywhere on the track, to mark a stretch of time, and it is highlighted with its length. Then you can make that bit its own piece, delete it, or apply an effect to it. You are not forced to work on whole pieces someone else decided the boundaries of.

Effects, in plain words

Under the Cut tab is a list of things you can do to a piece or a marked stretch, each with a line saying what it is for rather than what it is called: blur or pixelate the whole picture to hide a screen or a face, sharpen soft footage, clean up grain, darken the corners, black and white, warmer, cooler, a slow zoom in, camera shake for a hit, a soft glow, an old video look, or flip it left to right. They stack, they can be taken off again, and they survive cuts and saves.

Holes in the timeline

Pieces do not have to be jammed against each other. Drag one along the track and it leaves a hole behind, which plays as black with whatever music or text you have laid over it carrying straight on. Deleting a piece leaves a hole too, so you can drop something else into the space; Delete and close up is there when you would rather everything shuffled back, and Close the gaps tidies the whole thing up at once.

One thing to know: transitions and holes do not mix. If an edit has a hole in it the pieces are joined end to end and any transitions stand down, because a transition blends two pieces into each other and across a hole there is nothing to blend from.

Working on the timeline

The video starts as one piece, and there are five tools for turning it into an edit. You hold one at a time and it stays in your hand, so a run of cuts costs one click each. Escape puts whatever you are holding down and goes back to Select.

• Select (V) picks the piece you want to color or caption, and scrubs the playhead.
• Razor (C) cuts a piece in two where you click it.
• Trim (T) drags the start or end of a piece to change where it begins or ends.
• Erase (X) deletes the piece you click, leaving a hole where it was. Hold shift to close the hole up instead.
• Move (M) drags a piece along the track. So does dragging one with Select.

The timeline shortens as you go, so the preview always plays your edit rather than the original. Trimming pulls everything after it earlier rather than leaving a hole, because the pieces are simply laid end to end.

Cutting to the music

Add a music track and press Find the beat. It works out the tempo and lays a grid over the timeline, marked on the ruler with every fourth beat taller because bars are what an edit actually cuts to. From then on cuts, trims and layers all snap to that grid, so a cut lands on the beat without you having to aim at it.

Two things then become one click. Cut on every beat chops the selected piece at each beat inside it. Fit to the beat changes that piece's speed so it ends exactly on the next beat, which is the trick behind edits that feel locked to the track: nothing is trimmed away, the piece just plays a little faster or slower to land where the music does.

Speed is available on its own too, from a quarter to four times, and the sound is stretched with the picture rather than going out of sync. The timeline shows how much footage a piece holds and how long it takes on screen.

Sound you can see

Every sound has its waveform drawn into its block, including the video's own, so you can see where the hits are instead of guessing. The video's sound has its own controls: volume, tone, echo, or silence it completely. Give it its own track moves it onto a normal sound track, after which everything that applies to music applies to it, including animated volume.

Finding the moment you want

Press anywhere on the track and drag: the playhead follows your pointer and the picture updates as you go, so you scrub to a moment in one gesture rather than clicking and checking. A ruler above the track shows where you are.

For anything precise, use the keyboard. The arrow keys step a frame at a time, hold shift and they step a second. Up and down jump straight to the next cut. Then S cuts exactly where the playhead is, which is the reliable way to make a cut land: on a long recording a single pixel of track can be several seconds, so aiming with the mouse is hopeless and aiming with the playhead is not. The razor snaps to the playhead and to cuts you have already made, for the same reason.

P plays just the piece you are on and stops at its end, so you can cut and check, cut and check, without ever losing your place. Double-clicking a piece does the same. Space plays the whole edit, and Delete drops the piece you are on.

Nothing is destroyed while you work. The pieces are just ranges of the source file, so deleting a section costs nothing and changing your mind costs nothing.

Effects belong to a piece, not to the video

Color, framing and text are set per piece. Once you have split the timeline, the controls marked piece 2 are editing whichever piece is selected, and only that one. So you can brighten a dark opening, drop a black and white shot in the middle, and put a different line of text on each part, in one export.

Selection follows the playhead, so what you are watching is always what you are editing. Splitting a piece copies its look to both halves, which then go their own way. The frame shape and the quality setting stay the same for the whole export, because the parts are stitched back together at the end and that only works if they were encoded the same way.

Large recordings

The file is mounted from your disk rather than read into the tab, so a multi-gigabyte recording opens immediately. Only the pieces you keep are ever encoded.

Layers: text, images, another video, music

Everything you add gets a track of its own under the video, stacked up and lined up on the same ruler: text, a logo or sticker, a second video over the first, a music bed. You can see at a glance what is on screen when.

Everything lands at the playhead. Drag its block along its track to move it in time, drag the block's edge to change how long it stays, and drag the thing itself around on the picture to place it. Each track can be given a name of its own, which matters once there are a dozen of them. Shift-click anything on a track and it jumps so it begins at the playhead, which is the quick way to line something up with a moment: that works on layers, on caption lines and on keyframes.

What you see is what gets burned in.

Fonts and styles

Pick a ready-made look and it sets everything at once: plain white, punchy with a thick outline, yellow like broadcast captions, on a dark box, a soft shadow with no outline, italic and spaced out, or hot pink. Then adjust any of it: color of the words and of the outline, outline width, shadow, letter spacing, stretch, bold, italic, underline, and whether it sits on a box.

One typeface is built in. For anything else, use my own font takes a .ttf or .otf off your machine: it is used for the preview and burned into the export, so the file you get is set in your font, not an approximation of it.

Text that moves

A text layer can come in with a fade, a pop, a slide from any direction or a spin, and leave the same way. Pick the entrance and it plays in the preview immediately.

Selecting a layer puts a box around it with corner handles and a turn handle. Drag the middle to move it, a corner to resize it, the handle above to rotate it. That works on text, images and video alike.

Selecting a layer opens four more rows under its track: Position, Scale, Turn and Opacity. Each one holds that property's points, sitting on the timeline at the moment they happen. Click anywhere on a row to add a point there, drag a point sideways to re-time it, double-click it to remove it, or use the + at the end of the row to add one at the playhead.

The other way round is to switch on Animate and just work: with it on, every move, resize and turn records that moment as a point. With it off, the same changes move the whole layer instead. There is no limit to how many points a property has.

Adding your first point partway through does the sensible thing: it also drops one at the start holding whatever the layer already looked like, so the change runs from how it was to how you set it rather than applying to the whole thing at once.

Easing belongs to each point, not to the whole row. Click a point and it lights up; then choose how the movement leaving that point is shaped: Smooth (the default, eased at both ends), Steady, Ease in, Ease out, Overshoot, or Hold then jump. So one leg of a move can accelerate away and the next can overshoot and settle. With no point selected the picker sets the whole property at once, which is quicker when you want them all the same.

A selected point also gets its own panel: what it is worth, how it leaves, and a button to remove just that one. Delete removes the selected point rather than the piece of video.

A layer's path is drawn over the picture with its points marked on it, so you can see the shape of the movement and correct it.

The preview and the exported file read the same list of points, so the finished video moves exactly the way the preview did.

Blurring faces

Mark a stretch, or pick a piece, and press Find and blur faces. It looks through the footage three times a second, finds the faces, follows each one from sample to sample, and moves a blur with it in between. You can pixelate instead if you prefer that look, and it shows in the preview straight away rather than only in the finished file.

Sampling and following is what makes this possible in a browser at all: running a detector on every frame of a video would take longer than the video. Nothing is uploaded, the face finder runs in the tab like everything else, and it is about 43 MB the first time.

Speaking a line

Add a spoken line puts a voice track on the timeline. Type what it should say, choose from eighteen voices, set how fast it talks, and press Say it. The block sizes itself to however long the speech turns out.

Nothing about it is final. Click it later, change the words or pick a different voice, and say it again. The audio is regenerated from what you typed, so a spoken line is always editable, and a saved edit keeps the words rather than the sound: it is a few bytes instead of a megabyte and it can always be said again.

Sound, and what you can do to it

Music and sound effects each get their own track. Drop as many as you like: a bed of music underneath and a hit on a cut are just two sound layers.

Each one has a volume, a movement style for that volume, a fade in and a fade out, a speed, an echo (room, hall, canyon) and a tone: leave it alone, or make it muffled like it is behind a wall, thin like a phone, warmer, or brighter. Volume animates the same way everything else does, so a music bed can duck under a line and come back up after it.

One thing to know: the browser cannot play louder than 100%, so a volume above that is only audible in the exported file.

Captions you can move

Add captions listens to the speech and lays every line down as its own text layer. The button tells you what it is doing while it works, so there is no need to press it twice: it reports the model download, then that it is listening, then hands you the lines. Beside it is a choice of whether the lines cover the whole edit or only one piece of it, so there is nothing to guess at.

Captions arrive many lines at a time, so they fold into a single row that says how many there are. Open it when you want to work on individual lines and fold it away when you are done, and the rest of the edit stays readable either way. Ordinary text layers you add yourself always get their own row alongside.

Once they exist they are ordinary layers: drag one somewhere else, restyle it, retime it, animate it, or delete a line you did not want. There is a button to give every caption the look of whichever one you are editing. Lines that fall in footage you cut out are dropped, and a line that straddles a cut comes back as two, so nothing arrives at the wrong moment.

The speech model is about 145 MB and downloads the first time you ask for captions. It runs in the tab like everything else here.

Transitions

Click the badge on any join between two pieces to choose how one becomes the other: dissolve, flash, zoom, whip left or right, slide, circle, wipe, radial, pixelize, blur, diagonal, squeeze, through black. Every one of them was tested against the FFmpeg build this site actually ships, so none of them are decorative options that fail at export.

Transitions play in the preview, not only in the export: scrub across a join and you see the two pieces mixing, sliding or flashing into each other exactly where it will happen.

A transition overlaps the two pieces either side, so it makes the edit slightly shorter, and it cannot be longer than the shorter piece can spare. The timeline shows the pieces overlapping and the new total length as soon as you pick one.

Coming back to it later

Save this edit downloads the whole thing as a small file: every cut, layer, keyframe, style, transition and beat. The video and any music are not inside it, because a save with a music track embedded would be tens of megabytes and the files are already on your disk. Open an edit reads it back and then asks you to point at the files it needs by name. Everything else returns exactly as it was.

Burn one piece or the whole thing

Exporting is a real re-encode, so checking one cut by rendering the entire edit wastes minutes. The Export control lets you burn a single piece instead. Layers are re-timed onto that piece and anything that does not touch it is left out, so what comes back is that cut on its own, finished.

What it is not

One source video on the main track, one overlay lane and one sound lane. No nested edits, no color curves, no per-word caption styling. It covers the edit most short videos actually need: take the good parts, make them the right shape, grade each one, put moving text and a logo on it, lay music underneath, and join the pieces with a transition. For spoken captions, run the export through auto-captions afterwards.

Common questions

Related tools