Skip to main content

Editor overview

How the post-generation editor is laid out, what each panel does, and where to start when a scene looks off.

4 min read

Once the pipeline lands a draft, you're in the editor. It's the surface where most creators spend their hands-on time. This page maps the panels and the shortcuts.

The three panels

The editor opens with three regions: a script panel on the left, a scene timeline at the bottom, and a footage candidate grid on the right. Every surface in the editor is some combination of these.

Script panel. Each scene is a block. The block shows the script text, a small waveform preview of the voiceover for that scene, and a row of finding badges (issues the critic or vision evaluator flagged). Click into a scene to edit it.

Scene timeline. A horizontal strip at the bottom showing every scene with its current footage thumbnail. Click a scene to jump to it; the script panel scrolls to match. On mobile, the timeline collapses into a swipeable strip above the script.

Footage candidate grid. When a scene is selected, the right panel shows the candidates the orchestrator pulled, ranked by the vision evaluator. Each tile carries a confidence score, a provider chip (Pexels, Pixabay, NARA, etc.), and a click-to-preview affordance.

What the badges mean

Findings come from two sources: the script critic (Claude reviewing the script for blockers, majors, and minors) and the vision evaluator (Haiku rating each footage candidate for relevance + quality).

A blocker is a finding the pipeline considers serious enough to want addressed before publish, a defamation risk, a hyper-specific visual the footage can't deliver, a script section that violates the format's blocker rules. The sticky header at the top of the script panel shows the count.

A major is a finding worth looking at but doesn't block publish, wordy passage, repeated phrase, slightly off pacing.

A minor is a polish suggestion.

Findings have a resolve affordance. Marking a finding "looks good" or "apply suggestion" updates the script in-place. The sticky header counter updates live.

Keyboard shortcuts

The editor's keyboard layer is built around scene-by-scene reading. The full set:

ShortcutAction
jNext scene
kPrevious scene
1 through 9Jump to scene by number
+ EnterRegenerate the current scene with a hint
+ Shift + AAccept all blockers (use sparingly)
?Open the shortcut help sheet

The help sheet shows which shortcuts are active in the current context (some are scene-scoped, some are timeline-scoped) and is reachable from anywhere by pressing ?.

Regenerating a scene

The single most useful affordance after the first draft lands. Two flavours:

Regenerate scene, the script for this scene is rewritten from scratch using the brief + the outline. Quick when the issue is "the scene is fine in theory but the prose is flat".

Regenerate scene with hint, same as above but with a one-to-three sentence directive you write. "Less encyclopaedic, more dramatic." "Lead with the date, not the cause." "Cut the third paragraph entirely." The hint is treated as a directive Claude must satisfy, not optional advice.

The brief form's prompt coach has a Sharpen with Claude button that polishes a draft hint before you submit it, with the same niche + anti-clickbait guardrails the generation path uses.

Swapping footage

Each scene's footage is one click away from a different candidate. The candidate grid surfaces the orchestrator's full ranked list, and the per-scene Search more action runs an ad-hoc fresh search with a query of your choice.

Search more has a source-tier toggle: All / Archival / Free stock / Pixabay. The default is All, which walks the niche's default chain. Restricting to a single tier is useful when you know the niche should be archival-only (history, war, vintage) but the auto-pick included a stock clip you don't want.

If a tier-restricted search comes back empty, the input surfaces a one-click "Try All" affordance that re-runs the search across the niche default chain. You don't lose your query.

Custom voiceover at the editor stage

You can swap to a custom voiceover after generation. The voice editor accepts an MP3 / WAV / M4A upload and aligns the existing clips against your audio (rather than re-synthesising via ElevenLabs). A small waveform and an alignment indicator show whether your recording is on-target, light, or over for the script length.

What's next

FAQ covers the questions creators ask most about editor behaviour, including why a clip might not select, how to recover from a render error, and what happens to the timeline when you regenerate a scene.

Cadence is worth a read if the dashboard's editorial advisor surfaced after you started shipping at pace.

Cheers,
Carl