REA / VIDEO PRO MAX NATIVE / V0.1.1

AN INDEPENDENT CLAUDE / CODEX SKILL

Read the edit.
Make it yours.

A reference is full of decisions. Turn the supported ones into evidence, a reusable style, and a different video with your own words and media.

Inspect the actual run
01 / REFERENCE02 / EDIT GRAMMAR03 / NEW CONTENT04 / YOUR VIDEO

THE RECORDED HANDOFF

Same editing language.
Another story.

Recorded local run
Owned synthetic media · no live inference

A

The reference

OBSERVE / INFER
B

New content, adapted

COMPOSE / RENDER
0%

Linked playback compares relative progress across different durations. It does not claim frame identity or audio alignment. Use each player's controls to inspect independently.

REFERENCE SHOTS—observed intervals
SUBTITLE CANDIDATES—coherent OCR candidates, not speech claims
NEW VIDEO DURATION—from resolved project
VERIFICATION RECORD—read actual checks below

Inspect the timeline

SELECT AN EVENT

Loading recorded artifacts…

Each lane uses its own video's duration. A is the observed/inferred reference; B is the new resolved composition.

OPEN A RECORD

Reading local recorded evidence.

There is no upload form or live model behind this page. If an artifact is missing, it stays unavailable. Technical checks are bounded; the validation record describes the actual run and remaining gaps.

A USEFUL SYSTEM, NOT JUST A LOOK

Keep the decisions.
Change the content.

Observation, inference, reusable rules and your final composition stay separate.

01

Read what is there.

Inspect shot timing, visible words, subtitle appearance, coarse motion and signal cues. Keep timestamp evidence and the limits of coverage.

Analysis JSON ↗
02

Extract the grammar.

Normalize supported caption, layout, entrance and rhythm decisions. Keep an explicit record of inferred choices and unknown originals.

Style template ↗
03

Bind another story.

Add new text and independent assets. Reflow and retime the new words. The reference is never an implicit footage or music fallback.

Content bindings ↗
04

Make a playable result.

Render the supported 2D edit locally. Check the new content, style contract and decoded playback. Keep a portable project with asset hashes.

Native project ↗

YOUR NEXT REFERENCE

Give your agent
an editing vocabulary.

Use the skill, local CLI or optional workspace-bound stdio MCP. The native pipeline needs no paid model API.

Get the skill + quickstart
TRY A NATURAL REQUEST

“Analyze this reference's subtitles, rhythm and motion. Then use my new script and assets to make a different video in its style. Keep the template, timeline, project and verification.”

CLAUDE CODE / CODEX

Evidence before certainty.

A flattened video cannot uniquely reveal its hidden layers, original fonts, source clips, plugins or editing project. This release handles observable evidence and a bounded native 2D style vocabulary. Exact font recovery, arbitrary 3D/VFX, HDR and unrestricted NLE imports remain outside its verified scope. New words receive new timing; estimated timing is not ASR alignment.

Support matrix ↗Privacy ↗Measured validation ↗