Editing
The Edit sub-mode is Kitlig's timeline: a player, a media bin, and multi-track lanes for video and audio. It's the same document Storyboard, Assistant Editor, and Redub all work on, so a shot generated in Storyboard or a cut drafted by the Assistant Editor lands here as real clips.
The player
The player has no single video element sitting behind it. Two hidden video elements swap under the playhead as playback crosses from one source clip to another, so intercutting footage from different files never black-flashes between cuts. Audio plays through its own mix graph, not the video elements directly. Where no video clip sits under the playhead, in a gap between clips or after the last one, the player shows black, the same as the export does. A one-line subtitle shows the current clip's transcript, resolved from whichever analyzed sequence that footage originally came from, so subtitles keep working even for a clip dragged in from another sequence.
Five capture buttons sit at the right end of the transport bar. Copy frame to clipboard and Save Screenshot grab the frame as it appears in the player, so a quick still is one click away. The other three decode the exact frame from the clip's source file with ffmpeg and register it as an image asset in the project, with lineage back to that source, the same way the Source Monitor's Screenshot button works. Screenshot to project stops there. Edit Frame also selects the new asset in Sandbox mode, where the image operations and AI edits live. Open Frame in Canvas opens it as a new layered canvas for painting, masks and adjustment layers. Park the playhead over a video clip first; a gap has nothing to capture.
Media Bin
The left panel lists every video, audio and image asset in the project, with a search box and all / video / audio / image / generators filter tabs. Import brings a file in from disk. Drag an asset onto a lane: video and images go on video lanes, audio on audio lanes. Double-click a video or audio asset to open it in the Source Monitor and mark a range first.
Describing footage with AI
Every image and video can carry a description and search tags, written by a vision model: right-click an item and choose Analyze with AI, or press the Analyze (N) button to describe everything in view that has none yet. A video is shown to the model as a strip of frames with their times, plus a local transcript when the clip has speech, so besides the description and tags it gets a list of timestamped moments ("0:00–0:04 wide of the kitchen, 0:04–0:09 close-up of hands plating"). The tags show under each item, the search box matches descriptions and tags, and the Clip panel shows the same text for the selected clip, where you can edit it by hand (an edit survives a re-analysis). The editing agents read these too: that is how "find me the shot of the plate landing" works without anyone watching the footage. Each analysis is one paid call to the model set in Settings ▸ Agents ▸ Asset Analysis; the same panel can describe every new import automatically.
Image, colour and text clips
A still image dropped on a video lane becomes a clip that holds the picture for five seconds. Trim it to any length: a still has no end to run into. It exports like any other clip, letterboxed into the frame if its shape differs.
The Generators tab holds two clips that come from nothing. Drag a chip onto a video lane, or press its + to add it at the playhead on the selected clip's lane (or the first video lane).
- Colour clip: a solid colour the size of the frame. Pick the colour in the Clip panel. Useful under a title, as a slate, or as a background behind a scaled-down picture.
- Text clip: a title. The Clip panel's Text section has the words, the font, the size (as a share of the frame height, so it scales with the export), colour, alignment, bold and italic, line spacing, and an optional box behind the text. The player redraws the title as you type; the picture the export uses is written a moment after you stop.
- Title presets: lower thirds with one, two or three tiers (name, role, a subtitle line) on a translucent bar, the music-video stack (artist, song, album, bottom-left), a centred title and a boxed caption. Each is a chip in the Generators tab and a choice in the Text section's Preset picker. A preset title is made of lines, each with its own size, weight and colour (add or remove lines in the panel); Anchor places the block (bottom-left, top-centre…) with a margin from the edges, and Bar draws the band behind it. Once you have a look you like, type a name and Save as preset: it joins the Generators tab under My presets for this project.
- Timecode: a burnt-in clock. The player shows it live; Export Media draws one small image per frame at the export's rate and places it where the clip's anchor says. Timeline time counts from the sequence start, Clip time from the clip's own first frame.
None of these has sound, and the operations in the Clip panel (Augment, Lip Sync and the rest) work on footage only, so they stay hidden for stills, colours and titles.
Tools
Three mutually exclusive tools, shown as buttons on the toolbar:
- Select (V): move clips, or drag an edge to trim within the media's own bounds.
- Blade (B or C): click a clip to cut it at that point.
- Roll (N): drag the cut between two adjacent clips; both sides adjust up to their own media limits.
Split (S) cuts every clip under the playhead at once, on whichever tool you're using. A Blade click snaps the same way a drag does: within a few pixels of the playhead or of another clip's edge, the cut lands exactly there, so park the playhead where you want the cut and click near it. Delete removes the selection; select a gap instead of a clip and the button becomes a ripple delete, closing that gap and pulling everything after it earlier.
Bins
A bin is a named set of assets — a virtual folder, nothing moves on disk. The Bins row under the Media Bin's tabs has a chip per bin: click one to show only its assets, drop an asset card on a chip to add it, right-click a card to tick it into (or out of) any bin, and right-click a chip to rename, recolour or delete it. Bins are also how you point the assistant at material: "use the shots in the Food bin".
Editing with the Assistant
The Assistant (and an agent connected over MCP) has the same timeline in hand: it reads the lanes and clips, lays footage and titles, trims, moves, splits, adds markers, saves versions and exports. Its edits show up live, each one is an undo step, and each of its turns is saved as a minor version. See Assistant ▸ Editing the timeline by conversation. The toolbar's Reel button hands the sequence to the Reel Editor persona, which cuts a music-driven reel from a bin: Reel editing with the Assistant.
Subtitles
Right-click an audio lane (its header, or empty space on the lane) and choose Generate Subtitle Track…. Pick a style (a boxed caption at the bottom, a full-width lower band, or plain text), the characters per line (preset to what fits the sequence frame: a few words per line on a 9:16 reel, a full sentence on 16:9) and lines per cue, and press Generate. The job reads the speech of every clip on that lane: a clip's own transcript when it was imported through Redub, an earlier transcription of the same file, or a fresh local Whisper pass when there is none yet (set a Whisper model in Settings ▸ Local Inference first). The words are cut into readable cues, which land as ordinary text clips on a Subtitles video lane at the top of the timeline, one per cue, so you can retime, reword or delete any of them like a title. The lane also keeps an SRT of the cues as a text asset. Running it again on the same lane replaces the Subtitles lane rather than adding a second one, and Ctrl+Z removes it.
Versions
The toolbar's Versions menu (it shows the current number, v1.2, with a dot when the sequence has changed since) keeps snapshots of the sequence, like commits. Type what changed and Save makes the next major version (1.2 → 2.0); Minor makes a small step (1.2 → 1.3), which is also what the assistant does on its own edits. The list shows every version with its author, date, note, clip count and length; Switch makes one the live sequence again (your unsaved changes are kept as an automatic minor version first, so nothing is lost) and Ctrl+Z undoes the switch. Versions live in the project folder as plain JSON next to the sequence, so a project kept under git diffs cleanly; Kitlig itself never runs git.
Selecting several clips
Drag across empty lane space with the Select tool to draw a box; every clip it touches, on any lane, joins the selection (hold Shift to add to what is already selected). Ctrl+A selects every clip on the timeline, Ctrl-click adds or removes one, and Esc clears the lot. The selection moves as a group: drag any member and they all follow, including across lanes, and the lanes you are heading for show a dashed outline where each clip will land — brass when the lane takes it, red when it is the wrong kind (a picture clip over an audio lane), in which case that clip keeps its own lane at the new time. Delete, copy, paste, duplicate and the Clip panel's Transform all act on the whole selection, and one Ctrl+Z undoes the lot.
Markers
Press M to drop a marker at the playhead, or double-click the ruler. Markers show as coloured flags on the ruler with a hairline down through the lanes; drag a flag to move it, click it to select it, and Shift+, / Shift+. jump to the previous / next one. The Markers panel (toolbar button) lists them in time order with their timecode; click a row to park the playhead there and edit the name, note and colour below, or press Delete to remove it. Export ▾ in the panel writes the list as CSV (time, timecode, duration, name, note, colour; the times are timeline seconds) or JSON, and Export ▾ FCP XML… carries them as sequence markers.
Right-click a video clip whose media was described with AI and choose Add Scene Markers to mark its scenes in one go: a blue range marker per scene the clip shows, named Scene 1, Scene 2 and so on, with what happens there as the note. A long video uses the scene changes found while it was described; a short one uses its timestamped moments. Running it again adds nothing twice.
Copy, paste and duplicate
Clips copy and paste within the app. Ctrl+C copies the selected clip (or the Ctrl+click set) together with its linked audio or video partner; Ctrl+V pastes the copies with the first one at the playhead, keeping their spacing, on the lanes they came from. Ctrl+D puts a copy right after the selected clip. Hold Ctrl while dragging a clip and a copy follows the pointer while the original stays put; drop it anywhere on the same lane or another lane of the same kind. The clip's right-click menu has the same three entries. Pasted clips land on top of anything already there, so they can overlap; the pasted clip is selected, ready to nudge or delete.
Copying a look to other clips
Right-click a clip and choose Copy Attributes to remember its look. Then select the clips you want to match (drag a box, or Ctrl-click them), right-click one of them and choose Paste Attributes…, or press Ctrl+Alt+V. A short list lets you tick what to bring across:
- Text style: font, size, colour, alignment, bold and italic, line spacing, the box and bar behind the text, and where the title sits in the frame. Each title keeps its own words. A lower third's line styles (the bigger name line, the smaller role line) go onto the matching lines of the other title.
- Colour grade and LUT: every slider in the Color section plus the LUT and its intensity.
- Transform, Blend mode and opacity, Fades and Clip volume.
Each row says how many of the selected clips it fits. A text style only lands on titles, a grade only on footage and stills, and a volume only on sound clips, so you can select a mix of clips and paste safely. One Ctrl+Z undoes the whole paste. Copying a single clip with Ctrl+C also remembers its look, so Ctrl+C then Ctrl+Alt+V works the way it does in Premiere.
Editing titles together works without the clipboard too. Select several text clips and open the Clip panel on one of them: the Text section reads "Text · 3 clips", and a change to the font, size, colour, box or position goes to all of them. Typing the words changes only the clip you're on. The Color section works the same way for footage.
Fades
Click a clip's left or right edge (the trim handle, without dragging) to select that edge: it lights up brass. Press Ctrl+F and a half-second fade appears on that side; press it again to remove it. With a whole clip selected, Ctrl+F fades both ends. A linked video and audio pair fade together; Ctrl+Alt+F fades just the clip you clicked. A fade shows as a darkened corner with a diagonal line and a small square handle at the top: drag the handle to make the fade longer or shorter (it snaps to frames). The clip's right-click menu has Fade In and Fade Out, and the Clip panel has exact In and Out lengths in seconds.
A picture fade fades to transparent, so a clip on an upper lane fades to whatever is below it, and the bottom lane fades to black. A sound fade ramps the clip's volume. Both play in the viewer and render the same in Export Media.
Color
The Clip panel's Color section grades footage and stills the way Lumetri Color or Lightroom's basic panel does: Temperature, Tint, Exposure (in stops), Contrast, Highlights, Shadows, Saturation, Vibrance and Fade (lifted blacks). Drag a slider and the viewer updates as you drag; double-click a value to reset that slider, or press Reset for the whole grade. LUT applies a .cube file from the project, with an Intensity slider; the import button brings a new one in. With several clips selected the change goes to all of them, each keeping its other settings.
The grade is baked into a colour lookup table that both the viewer and the export use, so the render looks like the viewer. HDR footage from cameras that record HLG (many Sony, Panasonic and phone clips) is graded after it is brought to normal brightness, the way the viewer shows it.
Transform
Every clip on a video lane has a Transform in the Clip panel: X and Y move the picture (as a share of the frame, so 50 % is half a frame), Scale sizes it (100 % fits the frame), Rotate turns it clockwise, and Reset puts it back. With the clip selected and the playhead over it, you can also drag the picture in the player to move it (Shift keeps one axis) and use the mouse wheel over the player to scale it.
Lanes stack: an upper lane sits over the lanes below it, in the player and in the export alike. A full-frame clip covers everything beneath; scale it down or move it aside and the lanes below show through around it, which is how a picture-in-picture or a title over footage is built. The letterbox bars of a clip whose shape differs from the frame are see-through as well.
Blend modes and opacity
The Clip panel's Compositing section sets how a picture clip mixes with the lanes below it. Pick a blend mode from the list and set Opacity as a percentage; Reset puts it back to Normal at 100 %. Opacity multiplies with the clip's fades, so a 50 % clip fading in climbs from nothing to half strength.
Put the overlay on a lane above your footage and choose the mode that suits it:
- Overlay or Soft Light for a film grain or texture plate shot on mid grey. The grey disappears and only the grain stays.
- Screen or Add (Linear Dodge) for light leaks, lens flares and dust shot on black. The black disappears and the light brightens the shot.
- Multiply for paper, dirt or vignettes shot on white. The white disappears and the darker parts darken the shot.
- Darken, Lighten, Color Dodge, Color Burn, Hard Light, Difference and Exclusion work as they do in Photoshop or After Effects.
The modes blend with whatever is below, so a blended clip on the bottom lane blends with black. Transparent parts of the clip (a scaled-down picture, the letterbox bars, a fade) leave the lanes below untouched. The viewer and Export Media use the same formulas, so the render matches what you see.
Note
Hue, Saturation, Color and Luminosity aren't offered on the timeline yet, only in Canvas. Copy Attributes carries the blend mode and opacity to other clips like any other part of a clip's look.
In and out points
Press I to set the in point at the playhead and O for the out point, or use the In and Out buttons in the toolbar. The range shows as a brass band on the ruler, the lanes outside it dim, and a chip in the toolbar reads the two times; its × clears them. Setting a side that would cross the other clears that other side. The range never limits playback or editing, it only tells Export Media what to render: the dialog gains an Export the in/out range only checkbox, on by default while a range is set.
Note
The range is saved with the sequence, so it's still there when you come back.
Keyboard shortcuts
| Shortcut | Action |
|---|---|
| V | Select tool |
| B / C | Blade tool |
| N | Roll tool |
| S | Split every clip at the playhead |
| Space | Play or pause |
| K | Pause |
| L | Play |
| I / O | Set the in / out point at the playhead |
| Q / W | Trim the selected clip's start / end to the playhead |
| [ / ] | Nudge the selected clip's start (Shift for ×10) |
| , / . | Jump to the previous / next cut |
| Home / End | Jump to the start / end of the timeline |
| Del / Backspace | Delete the selection (clips, a marker), or ripple-delete a selected gap |
| Ctrl+A | Select every clip |
| Esc | Clear the selection |
| M | Add a marker at the playhead |
| Shift+, / Shift+. | Jump to the previous / next marker |
| Ctrl+C / Ctrl+V | Copy the selected clip(s); paste them at the playhead |
| Ctrl+D | Duplicate the selected clip right after itself |
| Ctrl+Alt+V | Paste Attributes: pick which parts of the copied clip's look go onto the selected clips |
| Ctrl+F | Fade the selected clip edge in or out (or both ends of the selected clip); press again to remove |
| Ctrl+Alt+F | The same, without fading the linked audio or video partner |
| Ctrl + drag | Drag a copy of a clip |
These are scoped to Edit and are disabled while you're typing in a field.
Clip references and frame edits
Double-click a video clip to open its Clip panel. The References section holds the images that belong with that clip. Add Asset picks project images; each thumbnail opens in Sandbox mode when clicked, and the × removes it. The same list feeds the Augment operation's reference images.
Edit Frame with AI makes a new image from one of the clip's frames. Choose an image edit engine, write a prompt, pick Clip start or Playhead (the playhead has to sit over the clip), and run it. Kitlig pulls that exact frame from the clip's source file, sends it to the engine together with the clip's references, and the result joins the references with a sparkle badge. Cutting the clip afterwards keeps the images on every piece.
Note
A generated version plays in step with the clip it was made for. Cut or trim the clip after the version lands and each piece keeps playing its own stretch of the generated video, so the second half no longer restarts from the beginning.
Lip sync and extend
The Operations section of the Clip panel also has Lip Sync and Extend tabs.
Lip Sync re-animates the mouth in the clip to a new audio track and lands the result as a version, like Augment. With Use timeline mix ticked, Kitlig renders the audible audio lanes under the clip's range and syncs to that, so a dialogue line you have already cut in on an audio lane is all it needs; solo that lane first so the music bed stays out of it. Untick it to pick an audio file instead, for example a clean dialogue stem. The audio plays from the file's start and is padded with silence or cut to the clip's length.
Extend generates footage that continues past the clip's last frame. Write what happens next, pick an engine, and run it. The new footage lands as a new clip right after this one on the same lane, named after the clip with "(extended)"; when the engine also made sound, that goes on the nearest audio lane as a linked clip, the same as dropping a video. Clips after it do not move, so make room first if the lane is full there. The engines take at most 15 seconds of source, so a longer clip sends only its tail.
Overlapping clips
Two clips can overlap on the same video lane, for example when you drop a clip onto a lane that already has one there. The clip you placed last is on top, and that is what plays and what Export ▾ Media… renders; the clip underneath shows again where the top one ends. Mute the top clip to see the one below it without moving anything.
Frame size
The Frame button in the toolbar sets the sequence's output size: a preset such as 16:9 1920×1080 or 9:16 1080×1920, a custom width and height, or Match first clip, which takes the size from the first clip's footage (a phone clip shot upright counts as upright). The preview letterboxes to that shape, so a 9:16 edit looks on screen the way it will export, and clips that don't fit the frame get black bars on either side, in the preview and in the export alike.
Export
Export ▾ Media… renders the timeline to one video file. It runs from the start to the end of the last clip on any lane, so a music track that carries on after the last shot plays over black rather than being cut off; set an in/out range to export less. The Frame row uses the sequence's frame size unless you pick another one for this render. Every cut lands on the frame it belongs to: a clip whose generated version came back a little short holds its last frame rather than pulling everything after it earlier. A small dialog also asks for the codec (H.264, H.265, VP9, AV1 or lossless FFV1, the same choices as the Re-encode tool), a speed preset for H.264 and H.265, and an optional target bitrate; leave the bitrate empty for a quality-targeted encode. The file type follows the codec: mp4 for H.264, H.265 and AV1, webm for VP9, mkv for FFV1. The Colour row picks the delivery colour. Rec.709 SDR (the default) is the normal web and phone format and looks like the viewer: HLG camera footage is converted the same way the viewer shows it, instead of coming out flat and grey. Keep HDR (HLG) writes an HDR file for HDR screens, with any normal clips lifted into it. Either way the file is labelled with its colour space so players don't have to guess.
The audio is rendered by the same engine the viewer plays through, so track levels, the compressor, limiter, reverb, pan, clip fades and the Mixer's master level come out exactly as you hear them. Timelines longer than 20 minutes use ffmpeg's mixer instead, which follows the same settings closely but not to the sample.
Export ▾ FCP XML… writes a Final Cut Pro 7 interchange file (the format Premiere Pro, DaVinci Resolve, and Final Cut 7 all read), probing each distinct source file first, then asks where to save it. Image clips and text clips go across as stills with a transparent background (a title is its rendered picture, redrawn first so it matches its current look at the sequence's frame size); colour clips and timecode clips have no file to point at and are left out. Each Subtitles lane is also written as an .srt file next to the XML, which Resolve and Premiere import as real, editable captions. The export message says what went out and what was left out. Colour grades and fades stay in Kitlig: the XML format has no portable way to carry them, so grade and fade in the other app, or export media. A clip's Transform goes across as a Basic Motion effect (scale, rotation, position), the sequence frame sets the XML's sequence size, and markers come along as sequence markers.
Package
Package ▾ As Directory… and As Zip… bundle the FCP XML together with a copy of every media file the timeline references. A bare XML export points at absolute paths on your machine, which becomes a relinking problem the moment you hand it off. Package fixes that by carrying the media along. You pick the destination first; the copy runs as a job, since real footage can take a while.