Video editor & moments

The video editor turns any uploaded video (or audio file) into a workspace for clipping highlights. From a single page you can scrub a zoomable multi-track timeline, mark named time ranges called moments, split and snap clips, export each moment to a standalone MP4, and surface interesting segments automatically with highlight filters that match against the file’s transcript and audio-event detections.

Opening the editor

Open any video or audio file in your library browser and choose Editor from the right-click menu, or navigate directly to a file’s edit page. The editor remembers which folder you came from, so the Back button returns you exactly where you started.

The editor uses an NLE-style layout:

• Top bar — back navigation, the file’s name and duration, and the local-inference job buttons (Transcribe, Waveform, and — once a transcript exists — Detect audio).
• Stage (left) — a chrome-less video preview with a transport bar underneath: jump ±5s, ~1-frame stepping, play/pause, a live timecode, playback rate (0.25×–2×), loop-the-selected-moment, mute/volume, and fullscreen.
• Inspector (right) — a tabbed panel: Moments, Transcript, Highlights, and Audio. The active tab and the panel’s width (drag its left edge) are remembered between visits.
• Timeline (bottom, full width) — the ruler, draggable moment bars, a merged markers lane showing highlight-filter matches in their filter colors, and the audio waveform track.

On phones and small screens the same panels stack vertically — player, transport, timeline, then the inspector — and the page scrolls.

---

Moments

A moment is a named time range on a video. It has a start and end time (stored to millisecond precision) and an optional description. Each moment can be independently exported, downloaded, and shared.

Creating a moment

Press M (or N) to create a new moment at the current playhead position, then:

1. Press I to set the in-point (start) to the current playhead.
2. Scrub forward to where the highlight ends.
3. Press O to set the out-point (end).

You can also drag either edge of a moment bar on the timeline to resize it, or drag the bar body to shift it in time. With snapping enabled (the magnet button, or G), dragged edges snap onto other moments’ edges and the playhead.

A focused moment bar can also be nudged with the keyboard: ←/→ moves it about one frame, Shift+←/→ moves it a full second.

Splitting a moment

Park the playhead inside the selected moment and press S (or the scissors button in the timeline controls) to split it in two. The right-hand half gets a numbered name (Clip → Clip (2)) and becomes the selection. Splitting commits what you see — any unsaved drag edits to that moment are saved as part of the split.

Editing a moment

Click any moment in the list or on the timeline to select it. The inspector switches to the Moments tab and shows an edit form where you can:

• Rename the moment and add a description.
• Type exact start/end times in seconds, snap either edge to the playhead, or jump the player to either edge.
• Export, download, or share the moment.

While you drag moment bars, edits appear highlighted in orange on the timeline to indicate unsaved changes. Click Save changes to commit all pending edits in one batch (each saved moment automatically re-exports). Navigating away with unsaved drag edits asks for confirmation first.

Keyboard shortcuts

Key	Action
Space or K	Toggle playback
J / L or ← / →	Jump back / forward 5 seconds
, / .	Step back / forward ~1 frame
R	Loop the selected moment
M or N	Create a new moment
I	Set in-point to current playhead
O	Set out-point to current playhead
S	Split the selected moment at the playhead
Delete / Backspace	Delete the selected moment (with confirmation)
Z or + / X or -	Zoom in / out on the timeline
F	Zoom to fit
A / D	Scroll the timeline left / right
C	Center the timeline on the playhead
G	Toggle snapping
← / → on a focused bar	Nudge the moment ~1 frame (Shift = 1s)
?	Open the keyboard reference

Shortcuts are disabled while focus is inside a text field or while a dialog is open. Press the keyboard button in the editor top bar to open the full reference.

---

The timeline

The timeline is a zoomable, scrollable multi-track strip: a tick ruler on top, your moment bars beneath it, a merged markers lane (when highlight filters have matches), and the audio waveform track at the bottom.

Zoom and scroll

• Zoom range: 1x to 50x. Use Z/X, the zoom buttons, or hold Ctrl/Cmd and scroll the mouse wheel. Zoom preserves the position of whatever you’re looking at — the playhead stays anchored on screen. F (or the fit button) resets to the full file.
• Scroll: use A/D, trackpad, or the scroll wheel. C snaps back to center on the playhead.
• When zoomed in, the timeline auto-scrolls to keep the playhead visible during playback (but never while you’re mid-drag).
• Scrubbing: press and drag anywhere on the ruler or the waveform to scrub playback continuously.

Waveform

The waveform track renders the audio amplitude as a mirror-image bar chart. It draws only the visible region, so it stays fast even on long videos at deep zoom. Click or drag on the waveform to seek.

Note

The waveform must be generated before it appears. Click the Waveform button in the editor top bar to queue the generation job. This runs once per video and the result is stored for future visits.

Markers lane

Every highlight-filter match renders as a slim marker in the filter’s color on a single merged lane — hover for the filter name, time, and evidence; click to seek. The lane appears automatically once any filter has matches.

Moment status pills

When a moment bar is wide enough, a status pill appears showing whether its export clip is Not processed, Processing, Processed, or Failed. A progress ring animates during encoding.

---

Exporting and downloading clips

Export a moment

Click Reprocess in the moment edit form. Alcoves queues an encoding job that:

1. Cuts the source video to the exact start/end times.
2. Re-encodes to H.264 video and AAC audio, clamped to a maximum of 1080p, with a web-optimized MP4 container (faststart).
3. Stores the result so future downloads are instant.

Export progress and an estimated time remaining are shown in the edit form and the status pill while encoding runs.

Tip

If you edit a moment’s time range after exporting, the old export is automatically invalidated and the status resets — the edit form shows an Edited since export chip. Export again to get a clip matching the new boundaries.

Download a clip

Once a moment’s status shows ready, click Download to save the MP4 file. If the clip isn’t exported yet, Alcoves queues the export and — as long as you keep the editor open — starts the download automatically as soon as encoding finishes. The export job itself always survives navigation, so coming back later and clicking Download again is instant.

---

Public share links

You can share an exported moment with anyone, even people who don’t have an Alcoves account.

Note

Sharing must be enabled on the library before share links can be created. A library owner or admin can toggle this in library settings.

In the moment edit form, click Share to open the share link panel. From there you can:

• Create a new share link — generates a unique token URL at /s/<token>.
• Revoke any existing link to immediately invalidate it.

The public share page at /s/<token> includes an embedded video player and is optimized for link previews in social apps and messaging tools (OpenGraph and Twitter player tags are server-rendered).

The video, thumbnail, and metadata are served without authentication. Only moments that have been successfully exported appear as playable on the share page.

---

Highlight filters

Highlight filters automatically surface interesting segments by matching against the video’s transcript and audio-event detections. Results update instantly as data loads — the matching runs entirely in the browser with no extra round-trips. Matches appear both in the Highlights tab and as colored markers on the timeline.

Note

Highlight filters require a transcript and/or audio detections for the video. Use the Transcribe button in the editor top bar to generate the transcript, then Detect audio (which appears once transcription completes). See the audio detection & transcription page for details on how Alcoves runs these locally.

Writing a filter expression

Each filter has a name, a color, and an expression. The expression is a short DSL:

audio:Laughter:40, word:goal

• Comma (,) means OR — the filter matches segments where either condition is true.
• Ampersand (&) means AND — both conditions must occur close together in time.
• audio:Label matches an audio-event detection whose label contains the word. A bare word without a prefix also defaults to an audio match.
• word:foo, keyword:foo, or text:foo match a transcript cue containing the word.
• :NN after an audio term sets the minimum confidence score as a percentage. For example, audio:Laughter:40 requires at least 40% confidence. The default is 20%.
• Values with spaces can be quoted: word:"oh no".

Examples

Expression	Meaning
audio:Laughter	Any detected laughter at 20%+ confidence
audio:Laughter:50	Laughter at 50%+ confidence
audio:Cheering, audio:Applause	Cheering OR applause
audio:Laughter & word:goal	Laughter AND the word “goal” within 5 seconds of each other
word:"oh no", audio:Screaming	The phrase “oh no” OR screaming

Proximity matching

When you combine terms with &, Alcoves checks that both conditions occur within a configurable proximity window (default: 5 seconds) of each other. You can adjust the proximity per filter in the filter settings panel. This lets you write filters like “applause during or just after a score” without matching applause that happens far away from any relevant spoken word.

Viewing results

Matched segments appear as a list of clickable timestamps below the filter. Click any match to jump the player to that point. Each match shows the matched evidence (the audio label or the transcript text) and a confidence score.

Filters show aggregate stats — match count, average score, and peak score — at a glance so you can quickly see which ones are producing useful results.

Built-in presets

When no filters exist for a library yet, the panel offers a Load presets button. This creates seven ready-to-use filters:

• Laughter
• Screaming
• Cheering
• Gunshot
• Profanity
• Reactions
• Funny clip (a combination filter)

Presets are a starting point — you can edit, rename, or delete any of them after loading.

---

Transcript and audio tabs

The inspector includes two signal tabs alongside Moments and Highlights:

Transcript — shows the full speech transcript as time-aligned cues. You can search the transcript to find specific words, and the active cue highlights as the video plays. A “Top words” tab shows the most frequent words in the transcript. If no transcript exists yet, the tab offers to run the transcription job directly.

Audio — groups audio events by label (e.g., “Laughter”, “Applause”) and shows a timeline strip of every occurrence. Click any occurrence to seek to it. Audio detection requires a completed transcript first; the tab says so until one exists.

Both tabs are populated by the same AI jobs that power highlight filters. Run Transcribe and Detect audio from the editor top bar (or from the tabs’ own empty states) to generate them. These jobs run locally on your server using CPU-only inference — no data leaves your instance.