RUNE-1013: Story AI-Native Editor Proposed
Summary
Rune.Story's editor is an AI-native semantic video editor. The Story script remains the source of truth, and UI projections are built from action order, wait progress, scene snapshots, and completion-source state.
Primary UI Model
The primary editor uses two columns:
- Preview on the left, showing the scene at the current action boundary or wait progress.
- Action trace on the right, showing one vertical row per main-axis action.
Rows are color-coded by action type and can show badges for diagnostics, animations, media work, completion-source registration, and scene mutations. Each row maps to a source span in the Story script.
Cursor and Seeking
The action trace has a cursor representing the current replay boundary. Dragging the cursor between action rows replays or dry-runs the script to that boundary and updates the preview. Instant actions such as create, non-animated set, layout solving, and completion-source registration are seek points but not time-bearing UI spans.
Only wait rows expose internal time. A wait row renders as an embedded mini timeline with an estimated or resolved duration and a progress pointer. Dragging that pointer scrubs frames inside the wait interval. Fixed duration waits are known immediately; waits on named sources, tags, speech, audio, or done may be indeterminate until dry-run/provider preparation resolves the active sources.
Debugging Projection
The editor should expose the same facts the runtime uses:
- selected action index, action type, arguments, and source span;
- scene snapshot and scene diff at the current boundary;
- completion sources created by each action;
- wait target resolution, active sources, duration, and blocking diagnostics;
- current wait progress when the cursor is inside a wait row.
This makes debugging action-oriented: authors inspect what an action changed and what a wait is waiting for, rather than only scrubbing raw frames.
AI Context
AI assistants should receive structured editor context in addition to preview imagery. A minimal context payload includes selected action metadata, source span, cursor boundary, optional wait progress, scene diff, diagnostics, and active completion sources. This lets AI apply local script patches, explain blocked waits, adjust animation presets, or insert actions near the selected row without reverse-engineering a flattened video timeline.
Rationale
Story is not a PPT clone and not a Premiere-style absolute timeline. PPT-like editors make static visual state easy but do not naturally represent wait semantics, completion sources, media timing, or deterministic replay. Traditional video timelines are excellent for clip trimming, but they flatten semantic actions into time spans and weaken source mapping.
Rune.Story keeps the semantic script visible. Action order is the main axis; time appears inside wait rows because waits are where Story playback advances time. This preserves reviewable text, source-aware diagnostics, AI-friendly structure, and visual frame scrubbing at the points where it matters.