Online Video Editor

Why every other free video editor asks you to upload

The pattern is so universal that people assume it is the only way video can work on the web: you upload the file, a server somewhere does the work, and you download the result. That architecture is what produces every frustration attached to free online editors. A gigabyte of somebody else's footage is expensive to receive, store and process, so there is a size cap. Rendering and transcription cost money per minute, so captions are metered, exports are watermarked, and the useful features sit behind a subscription.

Video is also the file type people are least comfortable handing over. A screen recording of a client call. Footage of your children. An unreleased edit. A deposition. "Deleted after one hour" is a promise, and a considerably weaker one than never having sent the file at all.

This editor inverts the arrangement. A complete build of FFmpeg and a speech recognition model are loaded into the page and run on your own processor. There is no upload, so there is nothing to cap, nothing to meter, nothing to charge for and nothing to watermark. Your project autosaves in your own browser's storage, on this device.

A timeline, not a form

Most free video sites are a sequence of forms: upload, choose an option, download, repeat for the next change. This is an editor. Clips sit on a timeline where you drag them into order, pull an edge to trim, and split at the playhead. Titles, logos and picture-in-picture have their own lane, positioned in time by dragging and positioned in the frame by dragging them in the preview. Music and voice-over sit on an audio lane with visible waveforms. Everything is undoable, step by step, and a whole drag counts as one step.

• Transitions between shots — crossfade, dip to black, wipes, slides, circle, clock wipe, zoom, pixelate
• Per-clip speed from quarter speed to 8×, with sound kept at its natural pitch, and reverse
• Fit, fill or blur-fill every shot to the frame, so mixed vertical and horizontal footage looks intentional
• Colour looks, hand adjustments and professional .cube LUTs, per clip or across the whole edit
• Titles with outlines, boxes and entrance animations, in ten bundled fonts
• Change the whole edit from 16:9 to 9:16 or 1:1 in one click, and every title keeps its place

Captions made on your device, word by word

Captions are the feature most people pay a subscription for, usually by the minute, because the popular caption styles — where the word being spoken pops or lights up — need a timestamp for every single word. The editor gets those timestamps from NVIDIA's Parakeet speech model running inside the page. It listens to the edit's own sound rather than the music, recognises 25 European languages without being told which, and punctuates as it goes.

Pick a style — bold pop, highlight box, karaoke fill, one word at a time, or classic subtitles in a box — and adjust the font, colours, size and position. Fix any word by clicking its line. Translate the captions into 20 languages, also on-device. Then burn them into the video, add them as a subtitle track viewers switch on, or download SRT and WebVTT files for YouTube or another editor.

The edits that take an hour by hand

Some of the most valuable things a paid editor does are not creative at all; they are tedious. Remove silences listens to the speech and cuts every pause longer than the length you choose, keeping a sliver of air either side so the result sounds tight rather than chopped. Remove filler words cuts "um", "uh" and "erm" at the exact timestamps the speech model found them. Music added under a talking video ducks automatically, dipping while someone speaks and returning in the gaps. Clean up voice removes hum, hiss and room noise, and loudness normalisation brings the mix to the −14 LUFS YouTube and Spotify expect.

All of these are ordinary edits on the timeline. You can see each cut, adjust any of them, or undo them all in one step.

What it costs to run in a browser

It would be dishonest to present this as free of trade-offs. A browser cannot reach the dedicated video encoder in your graphics card — the specialised silicon that makes desktop exports fast — so the final render happens on the CPU. On a modern laptop a 1080p export typically runs at around real time or faster, using several cores at once; a 4K export of a long edit is genuinely slow. Editing, previewing and captioning are not held back by this, only the final render.

The speech model is about 400 MB, downloaded once and then cached by the browser. And the size of what you can edit is bounded by your device's memory rather than a server's policy: long 4K sources are best trimmed into sections.

Exports, and one honest omission

Export presets cover YouTube at 1080p and 4K, TikTok, Reels and Shorts, the Instagram feed, a small 720p file for the web and email, and a file sized to fit WhatsApp's 16 MB limit. Formats include H.264 and HEVC in MP4, WebM, ProRes 422 Proxy in a MOV for Premiere, Final Cut and Resolve, animated GIF, and audio-only MP3 or WAV.

VP9 — the more efficient WebM codec — is compiled into the engine but deliberately not offered. On any real clip it exhausts the memory WebAssembly makes available and either crashes or hangs. Offering a codec that fails most of the time would be worse than not offering it, so WebM is written with VP8. Every feature and every limit is written up in the Video Studio guide.