Video Studio — the complete guide

Video Studio is a timeline video editor that runs entirely in your browser. It does the work a video subscription is usually bought for — multi-clip editing, transitions, titles, word-by-word auto-captions, silence removal, music that ducks under speech, colour grading, 4K export — with one difference that shapes everything else: nothing is uploaded. A complete build of FFmpeg and an on-device speech model run inside the page, on your own processor, and your project autosaves in your own browser.

This page is the reference. If you would rather just start, open the editor and drop a video onto it.

The editor at a glance

  • Left — panels: Media, Text, Captions, Audio, Transitions and Smart tools.
  • Centre — the preview, with play controls, a frame-accurate time readout, safe-zone guides and a button that saves the current frame as a PNG.
  • Right — the inspector, which shows the settings of whatever is selected, or the project's own settings when nothing is.
  • Bottom — the timeline: titles and overlays, the main video track, audio, and captions, from top to bottom in the order they stack in the picture.

Starting a project

Drop videos, photos or music anywhere on the editor, or choose a frame shape first — 16:9 for YouTube, 9:16 for Reels, TikTok and Shorts, 1:1 or 4:5 for feeds. If the first clip you add is vertical and you have not picked a shape, the frame switches to 9:16 to match it. Files of any size are fine: they are read from disk as they are needed rather than loaded into memory, so a multi-gigabyte recording opens as quickly as a short one.

Files the browser itself cannot play — ProRes, AVI, WMV, some MKVs, HEVC in some browsers — get a small preview proxy made automatically in the background, exactly as desktop editors do. The proxy is only for watching while you edit; exports always use the original file.

Editing on the timeline

  • Reorder — drag a clip along the video track; clips always butt together.
  • Trim — select a clip and drag its left or right edge.
  • Split — put the playhead where you want the cut and press S.
  • Delete and duplicate — Delete and ⌘D.
  • Snap — edges snap to the playhead and to other clips; switch it off in the timeline toolbar.
  • Zoom — + and −, ⌘/Ctrl-scroll over the timeline, or ⇧Z to fit the whole edit.
  • Undo — ⌘Z, as many steps back as you need. A whole drag is one step.

Select a clip to change its speed (a quarter speed to 8×, with the sound kept at its natural pitch), play it backwards, rotate or mirror it, and decide how it fits the frame: fit shows the whole picture with bars, fill crops to cover the frame, and blur fill puts the whole picture over a blurred copy of itself — the usual way to present horizontal footage in a vertical video.

Transitions

Select any clip after the first and choose how it arrives: a crossfade, a dip to black or white, a dissolve, wipes and slides in four directions, a circle opening or closing, a clock wipe, a zoom, or pixelation. A transition overlaps the two clips, so it makes the edit slightly shorter; it is limited to half the length of the shorter clip, which guarantees two transitions on the same clip can never collide. "Use this transition on every cut" applies it across the edit. What the preview shows is pinned to what the renderer produces, direction for direction.

Titles and overlays

The Text panel has styles for a big title, a subtitle, a name-and-role lower third, bold social text, a quote, a call to action, a handwritten note and a chapter card. Click one to add it at the playhead, type in the inspector, then drag the text into place in the preview and pull its corner square to resize it. Each title has a font (ten are bundled; all cover Western and Central European languages, most cover Cyrillic, and Inter and Roboto cover Greek), colour, outline, shadow, background box, and an entrance and exit: fade, slide up, pop, zoom or typewriter.

Images added as overlays become logos, stickers or watermarks — "Make it a watermark for the whole video" sets one up in a click. Videos added as overlays become picture-in-picture, with their own volume and an optional green-screen key.

Titles and captions are drawn by the same code in the preview and in the exported file, with the same font files, so what you see is exactly what you get. Text in scripts the bundled fonts do not cover — Hindi, Arabic, Chinese, Japanese — is drawn with your device's own fonts, and still renders correctly.

Auto-captions

Open the Captions panel and press Generate captions. The editor mixes the edit's own sound — clips and any recorded voice-over, but not the music — and a speech recognition model (NVIDIA Parakeet TDT 0.6B v3) transcribes it on your device. The model is about 400 MB; it downloads once and is then cached by your browser. It recognises 25 European languages automatically and punctuates as it goes.

Because the transcript is taken from the finished edit rather than the source files, every word's timestamp is already correct for the timeline, whatever cutting and speed changes you have done. That per-word timing is what makes the caption styles possible:

  • Bold pop — big uppercase words; the one being spoken pops in colour.
  • Highlight box — a coloured box sits behind each word as it is said.
  • Karaoke — words fill with colour as they are spoken.
  • One word — a single large word at a time, for fast-paced shorts.
  • Classic subtitles and Minimal — whole lines, for talks and tutorials.

Click any line in the caption list to correct it; word timings are kept where the words still match. Captions can be translated on-device into 20 languages, including Hindi, Arabic, Chinese, Japanese and Korean — per-word highlighting cannot survive translation, since word order changes, so translated words are spread evenly across each line. Import an existing SRT or VTT file instead of transcribing, and export SRT, WebVTT or a plain transcript at any time. On export, captions can be burned in, added as a subtitle track, or left out.

Smart tools

Remove silences listens to the edit's speech and cuts every pause longer than the length you choose, keeping a little breathing room either side. Choose the sensitivity for your recording: "only true silence" for a quiet room, "aggressive" for a noisy one. Remove filler words cuts "um", "uh", "erm" and "hmm" at the timestamps the speech model found them, transcribing first if you have not; "like", "basically" and "so" are optional and only removed when they stand alone between pauses. Titles, audio and captions after every cut move with the footage, and a single undo restores everything.

Split at scene changes finds the hard cuts inside a long recording and splits the clip at each one, so a single exported video can be rearranged shot by shot. Reframe switches the whole edit to another shape in one click.

Audio

Music added from the Audio panel runs under the whole edit, loops if it is shorter, fades out at the end and ducks — it dips automatically while the edit's own sound is loud and swells back in the gaps. Record a voice-over straight from your microphone: the edit plays while you speak (muted, unless you choose to hear it through headphones) and the recording lands on the audio lane at the point you started.

Two mix settings apply on export. Clean up voice removes rumble, hum and room noise from clips. Normalise loudness brings the whole mix to −14 LUFS, the level YouTube and Spotify play at. A safety limiter always catches peaks where music and dialogue add up past full scale. Ducking, noise reduction and loudness are heard in the exported file; the preview plays the untreated sound.

Colour

Every clip has looks — vivid, cinematic, golden hour, cool, vintage, black and white, faded film — and hand controls for brightness, contrast, saturation, temperature and gamma, plus a vignette. The Smart tools panel applies a look to every clip at once. Load a professional 3D LUT in the .cube format and it is applied on export with tetrahedral interpolation. The preview approximates temperature and does not show gamma or LUTs; the export is exact.

Exporting

Press Export (⌘E) and choose where the video is going:

  • YouTube 1080p and 4K, TikTok · Reels · Shorts, the Instagram feed, and Website / email at 720p.
  • WhatsApp — sized to stay under the 16 MB limit. MP4, HEVC and WebM exports can all target an exact file size.
  • HEVC / H.265 — about half the size of H.264 on newer devices.
  • WebM, ProRes 422 Proxy for Premiere, Final Cut and Resolve, and GIF.
  • MP3 or WAV — the mixed soundtrack alone.

The export renders natively at the resolution you choose — a 4K export draws its titles at 4K, rather than enlarging a 1080p frame — and shows progress, speed and time remaining. When it finishes you can play the file before downloading it.

How fast is it, honestly?

Slower than a desktop editor. A browser cannot use the video encoder built into your graphics card, so rendering happens on the CPU. The editor uses the multi-threaded engine — roughly twice as fast as a single thread, and shown as a badge in the top bar — and on a modern laptop a 1080p export typically runs at around real time or faster. 4K is slow for anything long. Editing, previewing and captioning are not affected; only the final render is.

Autosave and your project

The project saves automatically a moment after every change, and imported media is stored once when it is added, all in your browser's own storage on this device. Close the tab and reopen the editor, and your project is waiting. If a file ever goes missing from storage, its clips are marked offline; select one and choose Locate file to point it at the file again. "Start a new project" in the project settings clears everything. Private windows, and clearing site data, remove the project.

Keyboard shortcuts

  • Space play/pause · J/K/L back 5s, pause, forward 5s
  • ←/→ one frame · ⇧←/⇧→ one second · Home/End
  • S split · Delete delete · ⌘D duplicate · T add a title
  • ⌘Z undo · ⇧⌘Z redo · ⌘E export · ? show all shortcuts

Quick tools

For a single job, the quick tools skip the timeline: compress to an exact size, convert, trim, crop, resize, rotate, change speed, remove or extract sound, make a GIF, grab frames, watermark and join clips — plus these professional ones:

  • Stabilise — cancels handheld shake, with a slight zoom to hide the moving edges.
  • Green screen — keys out a green or blue screen, with automatic spill removal, over an image, video, colour or transparency.
  • Colour grade — looks, adjustments and .cube LUTs.
  • Clean up audio — noise and hum removal and loudness normalisation, without re-encoding the picture.
  • Audiogram — a podcast clip as a video with a moving waveform, title and cover.
  • Blur an area — blur, pixelate or black out a region you draw on the video.
  • Reverse & loop — reverse, boomerang, or repeat.

Compress, convert, resize, remove sound, extract audio, colour grade and clean up audio also accept several files at once and return a single ZIP.

Formats, and one deliberate omission

The editor reads essentially anything FFmpeg does — MP4, MOV, WebM, MKV, AVI, WMV, FLV, 3GP, MTS — along with MP3, M4A, WAV, OGG and FLAC audio, and PNG, JPEG, WebP, GIF and iPhone HEIC photos. VP9 is compiled into the engine but is not offered for export: on any real clip it exhausts the memory WebAssembly makes available and either crashes or hangs, so WebM is written with VP8, which always finishes.

The two engines

FFmpeg is compiled here in two builds. The multi-threaded one needs a browser feature called SharedArrayBuffer, which is only available on a page that is cross-origin isolated. That is incompatible with third-party ad frames, so it is switched on only for the editor, which carries no ads. The single-purpose tool pages run the single-threaded build, which produces identical output more slowly. If the badge on the editor says "1 thread", your browser declined isolation — usually an extension.

Privacy, precisely

Your video is read from disk by the page, processed by WebAssembly on your processor, and written back to a file you download. The speech model, the fonts and the FFmpeg build are all served from this site rather than a third-party CDN, so no other party learns what you loaded. The project and media are stored in your browser on this device and nowhere else. Once the editor and, if you use captions, the speech model have loaded, you can disconnect from the internet and everything still works.

Ready? Open the editor, or start with auto-captions, silence removal or compressing a video.