Video and audio

Video and audio files become playable cards on the canvas, in any format. One your browser can't play, such as ProRes, MKV or AIFF, plays from a copy Clusters makes, and Download gives the original. Audio cards draw a real waveform decoded from the file, and only one plays at a time so a board full of takes never becomes a wall of noise. There is no size or length limit on any plan; large files upload in parts, and the only limit is your drive.

Updated October 24, 2026

Both play in place on the canvas — the clip is a card among the reference stills it belongs with, which is what a video reference board is. A video also opens in a full review player when you need to look at it frame by frame.

Video

A video card plays inline. Move it, resize it, group it and comment on it like anything else. At rest it shows its poster frame, a small play mark in the top right corner so it reads as a video and not a photo, and its length in the bottom corner. Videos inside a grid get the same mark. Hover it for its controls: play, one frame back and forward, a scrubber along the bottom edge that lands on exact frames, the timecode, mute, and ⤢ to open the review player. The controls stay a usable size at any zoom.

Clicking a clip you haven't selected selects it, like any card. Click it again to play or pause, or use the play button in its middle. Double-click it to open the review player on that frame. On a board you can only view, one click plays. Dragging the card never starts it.

The review player

⤢ on a video card, a double-click on it, Open in player in its right-click menu (on the board or in Files), or Open in player in Quick look opens the clip full size in a dark room, on the frame you were on. Click the picture to play or pause; double-click for full screen. In full screen the controls step aside while it plays and come back when the pointer moves; on a touch screen, a tap brings them back without pausing. When you close it, the card shows the frame you left on.

Under the picture, the timeline fills the width. Below it, from left to right:

• the timecode, the clip's length and its frame rate;
• the transport: set in, a frame back, play, a frame forward, set out;
• loop, speed, framing guides, volume and full screen.

• Frame by frame. ← and → (or , and .) step one frame, ⇧← and ⇧→ one second. Hold the step buttons to keep going. The timecode counts the frame actually on screen, in the clip's own frame rate. Until that rate is read, it counts at 24 and shows the rate as ≈24. A phone clip with a variable frame rate says VFR: its frame numbers are close, not exact.
• J, K, L. Play backward, stop, and play forward; press again to go faster — the speed shows over the picture. Hold K and tap J or L to inch a frame; keep holding to play slowly.
• The timeline. Hover it for a preview of that frame, its timecode, and how far it is from the playhead. Drag to scrub; hold ⇧ while dragging to move a frame at a time. + and − (or ⌥=, ⌥−, pinch, ⌘-scroll) zoom in around the playhead until every frame has its own tick, with seconds marked above. The strip above the timeline shows where you are in the whole clip: drag it to move along. ⌥0 fits the whole clip again. While playing zoomed in, the timeline follows the playhead.
• Jump to a frame. Click the timecode, or just start typing, and enter 01:00:10:12, a frame number like 240, seconds like 10.5s, or a nudge like +12 or -1s.
• Timecode, frames or seconds. The chip next to the length switches the readout. Source timecode counts from the start timecode the camera or editor wrote into the file. Clip timecode counts from 00:00:00:00. Click the length to show the time left instead. ⌘C copies the timecode.
• In, out and loop. I and O set an in and an out point. They show as brackets on the timeline that you can drag, and as a chip with the length between them. ⇧I and ⇧O go to them, and ⇧Space plays from in to out once. ⌥L loops between them, or the whole clip when neither is set. ⌥X clears them.
• Speed from a quarter to double, from its button or ⇧< and ⇧>. The speed is remembered, and so are the volume and how the timecode reads.
• Framing guides for 2.39, 2.00, 1.90, 1.85, 16:9, 4:3, 1:1 and 9:16, action and title safe, thirds and a centre cross. G steps through them.
• Look closer. Z zooms the picture from fit to 100% (one pixel of the file per pixel of your screen) to 200%; drag to look around. Z again fits it.
• Share a frame. ⇧⌘C, or Copy link to this frame in the picture's right-click menu, copies a link that opens this player paused on this frame, for anyone who can open the cluster.
• Versions. A clip with versions shows them along the top, in a menu there past four or on a narrow screen. Switching keeps the frame, in, out and loop. Compare puts a second version behind the one on show: \ or the A | B buttons flip between the two on the same frame.

Esc closes it. ? lists every key, and so does Player.

In Quick look a clip plays with lighter controls over its bottom edge: play, frame steps (, and .), J/K/L, mute (M), the scrubber with its preview, and Open in player. K pauses there; Space closes Quick look as it does for any file, and ← and → still move between files.

A file the browser can't decode, ProRes outside Safari for instance, plays from a copy Clusters makes of it. While the copy is being made the player says so, and plays it when it's ready. See Formats.

Autoplay and looping

Right-click a video card for Autoplay (muted) and Loop. Together they turn a short clip into the thing a GIF used to be on a reference board — a moving still you never have to press play on — while the file stays a real video you can scrub, download and hand to an editor.

Autoplay is muted, and not as a preference: every browser blocks an autoplaying clip that makes noise, and a board of self-starting soundtracks would fight the one-at-a-time rule audio cards follow. Unmute from the card's own controls when you want to hear it.

A clip only plays while it is on screen. Scroll it out of view and it pauses. On a board holding twenty takes, autoplaying all of them at once is what makes a canvas stutter, so this is the behaviour rather than a setting.

Autoplaying clips keep the same time for everyone. Open a board nobody else is on and its autoplaying clips all start from the beginning, together. Anyone who opens it while you're there lands on the frame you're seeing, and the clips stay in step from then on, quietly catching up if one stalls to load. A clip that loops goes round; one that doesn't plays through once and stops on its last frame at the same moment for everyone. Pause one from its controls and it stays paused for you alone; press play and it rejoins everyone else.

To line clips up against each other, give one a start point: the frame it shows when the board starts. Right-click an autoplaying clip and choose Start here to use the frame on screen (pause on the exact frame first if you like), or Start at… to type a time such as 0:12 or 1:05.5. A looping clip still plays the whole clip, just shifted; one that plays once begins at its start point. The start point is saved on the card, so everyone sees the same alignment, and Clear start point puts the clip back to its beginning.

What Clusters knows about a clip

Right-click a video and choose Info, or select it and press ⌘I. Clusters reads the file's own container — MP4, MOV and M4V — and reports what an editor would look for:

A clip is read as it uploads, from the file on your computer. One uploaded before this existed is read the first time someone opens Info on it, a few kilobytes from the start (and end) of the file rather than the whole clip, and what it finds is kept on the card. WebM and other containers report less.

There is no size or length limit on video, on any plan — a reference clip, a cut-down or a full-resolution master all land the same way, and very large files upload in parts. The only limit is your drive.

Audio

Audio cards do more than provide a play button:

• A real waveform, drawn from the file — so you can see where the loud part is before playing it
• Cover art, lifted from the file's own tag when it has one — or set by hand on any audio card
• Standard transport controls

Only one audio card plays at a time across the whole board. Starting a second stops the first. On a board holding a dozen takes of the same cue, this is the only behaviour that makes sense.

There is no size limit on audio either, on any plan.

Cover art

MP3, M4A and FLAC files can carry artwork inside them, and Clusters reads it as the file uploads — a released track arrives on the canvas looking like itself.

WAV and AIFF have no standard way to carry a picture, so a loop exported to either will not have one. Right-click any audio card for Set cover image to add one by hand; a cover you set that way is never overwritten.

Downloading audio

Hover an audio card and a download button appears next to the transport. It is there for anyone who can see the card, including a signed-out visitor on a shared or published link — the same rule images have always followed.

The file comes back under its original name, not the card's title. Renaming a card to something readable never costs you the extension, which is what makes the file openable.

To take several at once, open the full Files page (F), select the ones you want and press Download — one zip, original names intact. Zips are capped at 500 files or 500 MB, whichever comes first; past either, download in batches.

Tempo and key

Sample and loop packs are named by machine — SFL_120_Gmin_Loop_Piano.wav, Cymatics - Orchid Kick 3 - 140 BPM.wav — so the tempo and key are usually already written down. Clusters reads them out of the filename as the file uploads and puts them on the card, where you can sort and filter by them in Files.

Both fields are editable: click the tempo or the key under the waveform and type. What you type wins permanently and is never overwritten.

Clusters does not listen to the audio to work either one out. Key detection on a single-instrument loop is guesswork — a kick loop has no key, and a hi-hat loop will report one anyway — and a wrong key is worse than an empty one, because you will trust it and your track will clash. If the name carries nothing, the field stays blank until you fill it in.

When a card has no waveform

The waveform is decoded from the file as it uploads. Decoding holds the whole track in memory, so files over 25 MB or longer than 10 minutes are skipped — as is anything in a format your browser cannot decode.

Those cards show an even, flat strip instead. The file still uploads, still plays, and still seeks; there is simply no shape to draw. A flat strip means "we don't know what this looks like", and it is deliberately not a guessed-looking waveform.

Cards that were uploaded before waveforms existed get theirs filled in quietly in the background, a couple at a time, while you have the cluster open.

Formats

Any video or audio file plays. H.264 in an MP4 or MOV plays in every browser as it is, and so do MP3, AAC, FLAC and WAV. Browsers differ on the rest: Chrome shows ProRes as sound with no picture and won't open AVI, WMV or MPEG files, Safari won't open MKV, and Firefox doesn't play HEVC, which is what an iPhone records.

For those, Clusters makes a playable copy: H.264 video with AAC sound in an MP4, or AAC for sound alone, at up to 1920 pixels on its long side. It is made once, on a converter Clusters runs, when the file is uploaded or first opened, and from then on every player uses it: the card, Quick look, the review player and a shared link. A short clip takes seconds and a long one a few minutes; the player says it is making the copy while it does. HDR footage is brought to ordinary colour for the copy.

The copy is for watching. Download always gives the original, byte for byte, and the copy never counts toward your storage. When the file itself is deleted for good, its copy goes with it.

A file over 4 hours or 20 GB gets no copy: it plays where your browser can play the original, and downloads everywhere. So does a file the converter can't read, and the player says it can't be played.

A clip that arrives without a file type, as some systems send a .mov or .mkv, is still a video.

What this is not

Clusters plays media. It does not edit it: no trimming, no cutting, no adjusting audio levels. Do that in your editor and upload the version you want people to see.

The photo adjustments available on image cards have no video equivalent.

The review player plays the file as it is, or its playable copy when your browser can't play the file, at the copy's size. It does not manage colour or show HDR, it can't play sound while you scrub, and backward play steps through frames rather than running smoothly. Compared versions line up by time from their first frame, so two cuts with different heads won't match frame for frame.

Storage

Media counts against your storage quota like everything else — 5GB free, 1TB on Creator, with the meter in Settings → Plan & billing. Video is usually what fills it.

Machine-readable: /docs/files/video-and-audio.md · /llms.txt
What changed and when: /changelog