Play anything.
Edit everything.

▶ A serious player

Hardware-decoded playback with real A/V sync, subtitles, frame-exact seek and live scrubbing — powered by libmpv when present. mpv/VLC keyboard muscle-memory works out of the box.

♪ Music, visualized

Audio plays with cover art or gnarly visualizers — musical spectrum, spectrogram, vectorscope, waveform — rendered by the playback engine itself. Press V to cycle.

🖼 Images, instantly

Stills open in the same window — ultrawide screenshots, 8K photos, and SVG, rasterized crisply. Convert to PNG, JPEG or WebP with resize in two clicks.

⬇ Convert like HandBrake

H.264, H.265, AV1, VP9 with quality presets or custom CRF; instant lossless remux; audio extraction to MP3/FLAC/Opus — straight from the player, no editing required. And when you've made a cut, export the edit itself — Reel renders your trims, splits and multiple source files into one new video, on the GPU encoder where your machine has one. One-click destinations — YouTube, TikTok, Reels, Instagram, Facebook, X — pick the frame, fit and codec for you.

⏺ Capture built in

Screenshot the full screen, a region or a window. Record your screen with system audio — no OBS, no external tools: Reel speaks your desktop's portal directly. Captures open in Reel, ready to trim or convert.

✦ Captions, on your machine

One button. Reel listens to the speech and writes the captions itself — locally, with no account, no upload and no per-minute billing. The speech engine and model are fetched automatically the first time (~85 MB, once); after that it works on a plane. Cues follow your edit: a line that spans a cut appears in both halves, and speech you trimmed away captions nowhere. What you see in the preview is exactly what burns into the export.

>_ Scriptable — and agent-ready

Everything the editor does, the command line does too, with no window at all: reel add, reel captions, reel title add, reel render --preset tiktok. Every command speaks --json, and reel commands --json hands a machine the whole interface. Projects are plain JSON files. Read the guide →

4K edits like 720p

Heavy footage gets an automatic background editing proxy the preview scrubs instead, while exports always use the original — and HDR clips (HLG, HDR10) are detected and tone-mapped so they look right, not washed out. Nothing to configure, nothing to relink.

Reel's own render engine

Timeline exports are drawn by Reel's GPU compositor frame by frame — ffmpeg only encodes. What the preview shows is what renders, by construction. And because every frame is composed live, parameters can move: keyframe exposure, colour, zoom, pan and PiP placement, with linear, hold or eased curves — from the panel or reel keyframe.

Green screen, built in

Check a box, pick the colour, done: a chroma key with soft edges and automatic despill that previews live — in the picture-in-picture inset too — and renders identically. Plus roll, slip and slide trims, live scopes while you grade, markers that become real MP4 chapters, and one-click frame export of the composed edit.

⧉ Picture-in-picture

Drop a clip on the overlay track and drag it where you want it, right on the preview — the inset plays live while you place it. Position and size are fractions of the frame, so what you place is what renders, at any resolution.

Speed, without the drift

Any clip from a quarter speed to four times, with the audio pitched to match and the clip's slot on the timeline resized to suit — so the rest of your cut stays exactly where you put it.

∿ Waveforms on every clip

The audio is drawn right on the timeline, so you cut on a word instead of hunting for it. Peaks decode in the background and are cached per source — split, trim, move or duplicate a clip and the waveform is simply already there. Copy, paste and duplicate with Ctrl+C/V/D (pasting inserts, so it can never quietly eat what you already placed), and drop markers with Ctrl+M to jump between with Ctrl+←/→.

T Titles you drag into place

Add text, then drag it where you want it — right on the picture. Size, colour, bold, outline. Position is kept as a fraction of the frame, so a title composed against the preview lands in exactly the same spot whether you export at 720p or 4K.

Curves, captions you can fix, publish everywhere

Five-point tone curves per channel, baked with your LUT into one GPU sample. The transcribed caption list is editable in place. And one click queues a render for every platform at once — each at its own frame and loudness.

Power windows

Limit any grade to an ellipse or rectangle — feathered, invertible, and keyframable, so the window can follow a subject across the shot. The colour goes where you point it; the rest of the frame stays honest.

LUTs and stabilization

Grade through any .cube 3D LUT — GPU-sampled, previewed exactly as rendered, never colouring the letterbox bars. And one checkbox smooths camera shake with a cached two-pass analysis whose test measures the shake actually halving.

Transitions, loudness, multicam

A real transition library — crossfade, dip-to-black, wipes, slides — previewed live with the exact geometry the render uses. Social presets deliver the mix at −14 LUFS so platforms don't squash your audio. And reel align syncs two takes by their sound alone — multicam without clap sticks.

✂ One click removes the dead air

Tighten finds every silence in your edit, cuts it with breathing room around the words, and closes the timeline up behind it — the podcast jump-cut pass in one button (or reel tighten in a script). Undoable, like everything.

♪ Sound that behaves itself

A volume trim per clip, and a music bed under the whole edit that ducks under speech automatically — and you hear it all live while you cut: the editor mixes every clip, fade and the ducked bed in real time. No volume curves, no keyframes: Reel rides the level for you, and hands it back the moment the talking stops.

✂ An editor when you want it

Hit Edit and your file is on a multi-track timeline: trim by dragging clip edges, split at the playhead, move with snapping, undo/redo, and save the cut as a .reel project that reopens with a double-click. Compositing Give clips a look — exposure, contrast, saturation, fades — and see it live in the preview: the editor and the export share one formula, so what you watch is what renders. Crossfade between clips, reframe a landscape shot into a vertical one, and queue every platform at once.

Codecs — handled for you

Reel plays best with libmpv (hardware decode, audio sync, subtitles, visualizers) and ffmpeg (convert & record). On Arch the package installs both automatically; the one-line installer offers to install them anywhere else; and if ffmpeg is still missing, Reel downloads its own private copy on first launch. No codec hunting.

Your default player, if you want

On first launch Reel asks whether to open Video, Music and/or Images by default — tick what you want, done. From then on double-clicking a file opens it in Reel. Change your mind any time under ⚙ → Default apps; Reel always stays in your "Open with" menus either way.

First run

reel video.mkv     # play immediately
reel song.flac     # music + visualizer
reel photo.svg     # view images

The first screen recording shows your desktop's own "share screen" picker — approve it once and Reel remembers.

Convert a file

Open it → ⬇ Export → pick format & quality → Start. Live progress, cancellable, never overwrites. Works on video, audio and images alike.

Caption a video

Edit → Generate captions → wait a moment. The lines appear on the preview where they will burn in; drag the Size slider to taste, then export. Nothing leaves your machine.

Record your screen

Reel’s tray icon → Record → choose screen, window or region in the system picker → ⏹ Stop. The clip opens in Reel — trim it or convert it on the spot.

Grab a screenshot

Reel’s tray icon → Screenshot → Full screen, Region or Window. It opens in Reel immediately — export to JPEG/WebP or just find it in ~/Pictures/Reel.

The whole thing, readable

About ten thousand lines of Rust: the player, the editor, the export engine, the capture path and the command line. No telemetry, no accounts, no phone-home — captions run on your own machine precisely so nothing has to leave it.

github.com/Pixygon/Reel →

Contributing

Bugs and ideas are welcome as issues; the architecture notes that guide the code live in CLAUDE.md, and the roadmap is honest about what is and isn't done yet. Every feature ships with a test that proves it against real ffmpeg output.

Open an issue →

Support the work

Reel is free and will stay free. If it saved you a subscription, you can help keep it moving — and starring the repo genuinely helps people find it.

★ Star on GitHub