Video Tools

What these tools are for

Most video jobs are small edits, not full productions. You need to cut the dead air off the front of a clip, drop the file size so it will send, or turn a recording into something a website will accept. The Video Trimmer, Video Compressor and Video Converter cover that ground.

The rest handle shape and sound. Video Cropper and Reframe for Vertical Video square a wide clip up for a phone screen. Rotate and Flip a Video fixes footage that came off the camera sideways. Mute & Replace Audio swaps the sound. Frame Grabber pulls out a still for a thumbnail, and Video to GIF makes a short loop that plays anywhere.

Doing video work in a browser tab

Video is the biggest thing most people ever edit. In a normal online editor the slow part is not the edit at all: it is sending a large file up to a server and waiting for the result to come back down. That step does not exist here. The file is already on the machine doing the work, so you open it and start.

The trade-off is that your own computer does the encoding. A long clip at a high resolution takes time, and an older or busy machine takes longer. Browsers also vary in which video formats they can decode, so an unusual file may not open at all. Leave the tab visible while a job runs, since browsers slow down work in background tabs.

Getting a good result

Pick the tool that does the least. If you only want the middle thirty seconds, trim first and compress afterwards. Cutting before you re-encode leaves less video to process and a smaller file at the end. Screen & Webcam Recorder and Subtitle Tools are worth knowing about too, for capture and for captions.

The common mistake is re-encoding the same clip over and over. Every pass through a compressor throws detail away, and it never comes back. Go back to the original for each new version, and keep that original somewhere safe. It also helps to check what the destination needs, such as aspect ratio and length, before you edit rather than after.

Two export speeds, and what puts your clip on the fast one

There is no ffmpeg build behind these pages. The shared video code uses parts the browser already has: a hidden video element, a canvas, and the recorder built into the browser. That is why a video page here weighs about the same as any other page on the site. It is also why the output menu is a probe rather than a promise. The code asks the browser which formats it can really write, keeps the best working option for each file type, and hides the rest. On most machines that leaves MP4, WebM, or both.

The older way to write a file is to play the clip and record the canvas as it goes. Ten seconds of video costs about ten seconds of waiting. The Video Joiner, GIF to Video, the Photo Slideshow Maker and the Stop Motion Maker all still work this way. Each of them builds a picture that exists only while it plays, so there is nothing else to read.

Video Studio and the Video Splitter take a shorter route when they can. The file is taken apart, its frames are decoded by the browser itself, changed on a canvas, encoded again, and written into a fresh MP4. Nothing plays and nothing is audible, and the tab does not have to stay in front. Five things have to be true first. The browser must support WebCodecs. The file must really be an MP4 or a MOV, which is checked by reading the ftyp marker in its first bytes rather than by trusting the name. It must be 500 MB or smaller. MP4 must be the format asked for. And the sound must not be getting swapped for a live microphone.

Sound normally rides across that route untouched. The audio packets are lifted out and re-timed rather than encoded again, so nothing is lost and it costs no time. A speed change is the one case where that cannot work, and there the sound is decoded, replayed at the new rate, and encoded fresh, pitch and all. If anything at all trips on the fast route, the same job restarts on the slow one and the wait estimate changes to match. The file you get is named after what actually came out, because the fast route always writes MP4 whatever the menu said. Reading frames was never the slow part: the stills, the GIF and the contact sheet in Video to Pictures seek to each moment they need instead of playing anything.

Stacking edits so the encoder only runs once

Trim, crop, rotate, colour and the rest used to be separate pages. Doing three of them meant three downloads and three trips through an encoder. Each trip costs a little picture and a little sound, and it can slide the audio against the vision. Video Studio takes a list of steps instead and folds the whole list into one pass.

The steps come in three kinds, and the kind decides when each one runs. Trim, speed and sound are read before a single frame is decoded, because they pick which samples go in and what the file says about itself. Rotate and flip, crop, reframe and resize change the shape; each draws the last canvas onto the next, so every step is handed a whole picture. Colour, green screen, hiding a face or a plate, and the watermark are painted on each finished frame, in the order you arranged them.

The shape steps could have been fused into one transform, which would be quicker. They are kept apart on purpose. Straightening a tilted horizon can fill the empty corners with a blurred copy of the frame, and reframing can pad the sides the same way. Neither is a plain change of coordinates, so the chain of canvases stays. The painting steps work on what has already been drawn, not on the raw footage. Otherwise the blur over a number plate would punch a hole straight back to the ungraded original.

Resize is the step that also speaks for the encoder. Its presets pair a height with a bit rate, since neither number means much alone: 480p at 800 kbps, 720p at 1800, and 1080p at 4000. It only ever goes down. Asking 1080p of a 720p clip cannot put back detail that was never recorded. Sizes are rounded to an even number of pixels as well, because H.264 cannot encode an odd one and fails without saying so.

Telling apart the tools that sound alike

Several pages here overlap at the edges, and what separates them is usually what comes out rather than what goes in. Video Studio hands back one clip from the part you marked. The Video Splitter works out every cut first, shows you the plan, then exports the pieces one at a time and saves each as its own file. Its overlap setting reaches backwards, so each piece starts a little before the last one ended and nothing falls between them. The pieces are made strictly in turn, because two recordings from the same video at once give you two broken files instead of two good ones.

Video to Pictures is three outputs behind one menu: a GIF, a set of stills, or a contact sheet. The trim range applies to the first two. The sheet always spans the whole clip, and instead offers to start and stop two per cent in, because the first and last frames of a video are so often black. It samples the middle of each slice rather than the edge, so the frames look picked rather than caught. Stills come back as a ZIP. GIF to Video runs the other way, and it is the one job that cannot be hurried: an animating GIF will not tell the page which frame it is on, so it has to be watched in real time.

The Photo Slideshow Maker and the Stop Motion Maker both build video out of pictures, from opposite ends. The slideshow times photo files you already have. Stop motion takes one webcam frame at a time and lays a ghost of the last frame over the live view, so you can see how far the model moved. The Audiogram Maker starts from sound instead, and draws its waveform from the decoded samples rather than from a live meter, so the picture matches the audio frame for frame. Subtitle Tools is the only page here that finishes at once, because it never touches the video at all. SRT and WebVTT are the same idea with different punctuation.

Two limits are worth knowing before you start. Loop a Video has to decode a whole pass of frames before it can write the longer file, so the selection is held in memory while that happens. The number of repeats adds nothing to that, because the one pass is written out again and again rather than copied. It picks the largest frame size whose full run fits a 700 MB budget, and shrinks the picture rather than running out half way. The GIF builder warns as soon as your settings ask for more than 400 frames, and refuses past 1,500. That is the point where the file gets big and the wait gets long.