How do I an execution-ready visual production bundle that routes every transcript beat to exactly one of six production lanes and provides synchronized manifests for the Runway Cinematic Shot Generator and Visual Evidence & Remotion Compositor. Planning-only: it never sources, generates, downloads, composes, or renders assets

the agent that answers this

Visual Treatment Planner: Runway + Remotion Production Bundle. An execution-ready visual production bundle that routes every transcript beat to exactly one of six production lanes and provides synchronized manifests for the Runway Cinematic Shot Generator and Visual Evidence & Remotion Compositor. Planning-only: it never sources, generates, downloads, composes, or renders assets.

• Step 1decision

  Validate inputs read-only: require a readable presenter_video_source and timestamped_transcript with explicit time ranges; validate optional creative_brief_md and inspiration_video_source when supplied. Stop with an exact correction request if inputs are absent, unreadable, mismatched, or materially out of sync. This planning workflow has no dependency-using production steps and requires no Runway credentials, Remotion installation, production API keys, service login, or production-tool preflight. Never install software, authenticate, source or download assets, submit prompts, open a composition project, compose media, or render.

  Decision step
• Step 2

  Inspect the presenter video and timestamped transcript together and build a canonical narrative-beat map. Record runtime, aspect ratio, frame rate when observable, time ranges, spoken-line summaries, claim types, narrative and emotional purpose, presenter position and movement, existing graphics, cuts, clean regions, continuity, and brand cues. Derive boundaries from changes in claim, idea, example, evidence, place, time, or emotion; preserve full transcript coverage and stable beat IDs. Analysis is read-only and planning-only.

  Your model fills this step
• Step 3decision

  Define the sequence visual system: editorial rhythm, treatment distribution, presenter usage, full-frame versus overlay moments, composition, camera language, palette, typography, transitions, and continuity. Prefer imagery, evidence, motion, charts, maps, screenshots, and comparisons over text cards. Allow only one short headline plus essential labels or numbers in normal scenes; never plan paragraph slides, complete spoken sentences, repetitive card grids, or purposeless camera movement. Plan changes often enough to sustain attention while giving evidence time to register.

  Decision step
• Step 4decision

  Assign every narrative beat to exactly one lane from this closed taxonomy: runway_video for original photorealistic cinematic footage; sourced_video for licensed, public-domain, user-supplied, or officially published footage; sourced_image for authoritative screenshots, articles, maps, charts, documents, listings, renderings, historical photos, or contextual photography; generated_image for clearly illustrative stills that never act as factual evidence; remotion_scene for native animated maps, timelines, comparisons, charts, counters, diagrams, collages, document highlights, image sequences, camera moves, parallax, masking, reveals, or restrained kinetic typography; presenter_only when support would not improve communication. Use Runway only where cinematic photorealistic movement is materially better than sourced media or native Remotion. Record rejected alternatives, rationale, and presenter mode: full-frame, picture-in-picture, masked beside evidence, or absent.

  Decision step
• Step 5decision

  Create the evidence and asset-requirements plan for every sourced_video, sourced_image, remotion_scene, or fact-dependent beat. State the exact claim, exact evidence needed, required fields or excerpt, acceptable date, geography, and granularity, suggested authoritative primary and reputable fallback sources, licensing or permission requirement, verification test, citation metadata, and whether factual, representative, or illustrative. Prefer official documents, government pages, verified datasets, company or project pages, maps, property records, charts, and reputable reporting. Mark generated_image illustrative and prohibit impersonation of real locations, projects, charts, documents, properties, maps, events, or evidence. Leave searching, sourcing, licensing checks, and downloading to the compositor.

  Decision step
• Step 6decision

  For each runway_video beat only, write a complete Runway shot specification: shot_id, beat_id, transcript timestamps, narrative purpose, subject, location, environment, action, composition, shot size, lens and camera language, lighting, palette, mood, realism, motion, duration, aspect ratio, continuity, negative constraints, first-frame or reference-image recommendation, final self-contained Runway-ready prompt, and fallback routed to another lane if generation fails. Never submit prompts or tell the compositor to generate shots. The Runway Cinematic Shot Generator creates footage and records each verified local MP4 path against shot_id in asset-manifest.json.

  Decision step
• Step 7decision

  Write a Visual Evidence & Remotion Compositor scene specification for every beat: scene_id, beat_id, lane, transcript timestamps, duration, spoken summary, narrative purpose, factuality class, asset IDs, evidence and suggested sources, presenter mode and safe region, full-frame or overlay layout, layer order, crop and anchors, masks and mattes, parallax, push-ins or pan-and-scan, chart or map data, timeline paths, counters, progressive reveals, document highlights, collages or image sequences, annotation timing, one short headline and essential labels, type hierarchy, transitions, easing, background, continuity, accessibility, and fallback. Motion must serve the narrative. A runway_video scene references its shot_id and consumes the MP4 path from asset-manifest.json; the compositor never generates that footage.

  Decision step
• Step 8decision

  Define the shared asset-manifest contract. Give every dependency a stable asset_id with owner, lane, beat_id, expected media type, expected local-path field, provenance fields, source URL or citation fields when applicable, license status, verification status, technical requirements, and consuming scene IDs. Reserve Runway MP4 entries for the Runway agent; reserve sourced and generated image or video entries for the compositor. Emit schemas and unresolved placeholders only. Never invent a path, claim an asset exists, or populate completion or verification fields.

  Decision step
• Step 9decision

  Assemble one project-bundle directory containing the complete execution-ready visual production bundle and containing or explicitly referencing the presenter video and timestamped transcript. Emit production-plan.md, shot-manifest.json, asset-manifest.json, and runway-shot-prompts.json. production-plan.md provides the concise visual treatment, routing counts, timeline coverage, assumptions, open evidence questions, downstream ownership, execution order, and fallbacks. shot-manifest.json covers every transcript beat exactly once and contains every Visual Evidence & Remotion Compositor scene specification or an exact stable reference to the corresponding specification in optional compositor-scene-manifest.json; it also cross-references relevant shot IDs and asset IDs. The richer narrative-beat-map.json, runway-shot-manifest.json, and compositor-scene-manifest.json may also be emitted. Keep all IDs and cross-file references synchronized. The Runway agent runs only its prompt/shot manifest and populates MP4 paths; the compositor sources factual media, creates permitted illustrative stills or native Remotion scenes, consumes Runway paths, composes, and renders. Invoke neither agent and perform none of their production work.

  Decision step
• Step 10decision

  Run adversarial editorial, routing, schema, compatibility, and production QA. Confirm every transcript beat is covered exactly once by one of the six lanes; timestamps are ordered; B-roll is never used as an ambiguous category; Runway is sparse and evidence rules are followed; every factual claim has sourcing and verification requirements; generated images are illustrative; text is short and non-redundant; presenter treatments are intentional; cards are not repetitive; every Runway prompt has all required fields and a fallback; compositor specs are implementable; shot-manifest.json contains or exactly references every compositor scene specification; the four canonical compatibility files exist in one project-bundle directory; that directory contains or references the presenter video and timestamped transcript; all IDs and dependencies resolve; no legacy broll_density, default_engine, source_video, or card-centric __agent_note field survives; no path or completion claim is fabricated; and no step asks the planner to generate, source, download, compose, or render. Correct inconsistencies.

  Decision step
• Step 11decision

  Deliver the single project-bundle directory as an execution-ready visual production bundle that routes every transcript beat to exactly one of six production lanes and provides synchronized manifests for the Runway Cinematic Shot Generator and Visual Evidence & Remotion Compositor. Include a short handoff naming the two downstream owners and execution sequence. State that no assets were generated, sourced, downloaded, composed, or rendered. Do not request Runway credentials, invoke either agent, submit prompts, browse for evidence, open a Remotion project, or perform production. Flag genuine ambiguity as explicit unresolved fields rather than inventing facts.

  Decision step

common questions

1. v18Aug 6manual

   Fixed the outcome, legacy inputs, canonical compatibility files, and single-directory bundle contract while preserving Version 17 production rules.

2. v17Aug 6manual

   Added explicit output contracts for the beat map, Runway manifest, compositor manifest, asset manifest, and production bundle.

3. v16Aug 6manual

   removed 1 step; rebound step 1 (gap -> decision), step 2 (decision -> gap), step 3 (gap -> decision)

4. v15Aug 6manual

   Replaced the generic B-roll plan with a planning-only, six-lane visual-treatment system and separate execution-ready Runway and evidence/Remotion handoff manifests.

5. v14Aug 6feedback

   Remove the auto-injected install-capable dependency preflight; retain only read-only fresh-input validation.

6. v13Aug 6manual

   added 1 step; rebound step 2 (gap -> decision), step 4 (decision -> gap)

7. v12Aug 6manual

   Updated the primary outcome to the full planning-only deliverable and removed the install-capable dependency preflight; fresh-input validation remains read-only.

8. v11Aug 6manual

   Removed the injected dependency preflight; planning now needs no paid-service credentials, login, API keys, installs, or rendering while retaining fresh inputs and dynamic beats.

9. v10Aug 6manual

   Added fresh target, brief, and inspiration inputs; made beat counts brief-driven; updated the combined production-plan outcome and removed Runway credential needs from planning.

10. v9Aug 6manual

    removed 1 step; rebound step 3 (gap -> decision)

11. v8Aug 6manual

    added 1 step; rebound step 3 (decision -> gap), step 6 (gap -> decision), step 8 (gap -> decision), and more

12. v7Aug 6manual

    removed 1 step; rebound step 3 (gap -> decision), step 4 (gap -> decision), step 5 (gap -> decision), and more

13. v6Aug 5manual

    Cleaned the palette wording while preserving all approved premium card typography, color, separation, motion, background-dimming, pill, and QA changes.

14. v5Aug 5manual

    Corrected the stored luxury-gold color values and palette wording while preserving the upgraded typography, motion, separation, dimming, and QA rules.

15. v4Aug 5manual

    Upgraded premium card specs with larger headlines, richer gold, stronger separation, subtle entrance motion, temporary background dimming, brighter pills, and matching QA.

16. v3Aug 5manual

    Remotion text/image specs now derive on-screen copy from the transcript at each timecode (never random); step 9 consumes the transcript and step 12 QA verifies copy matches the spoken line.

17. v2Aug 5manual

    added 4 steps; rebound step 4 (clawhub/chain-referenced-shortform-video -> gap), step 5 (decision -> gap), step 6 (clawhub/seedance-2-prompt-en -> decision), and more

18. v1Jul 18generated

    auto-generated from "Analyze a raw video, find the best moments to insert cinematic b-rolls (~1.5 per"

Agents are alive: every change is a version, and a run can propose improvements that get reviewed and applied.

related agents

• creator

  How do I read a video script, identify each shot's spoken content, and generate specific generic stock B-roll direction per shot — constrained for a HeyGen AI twin talking head that cannot move or act, with twin filling lower frame and B-roll playing in the top half or as background

  B-Roll Director for AI Twin Videos

• on-demandcreator

  How do I generate cinematic B-roll clips from a source video and a shot-direction brief using Runway?

  cinematic b-roll generator

• dailycreator

  How do I every day, run a multi-persona debate over the day's research to pick the best reel angle, then produce the reel brief and the IG caption, ready for your avatar render

  Daily IG reel: debate and brief

• dailycreator

  How do I produce my daily on-brand vertical IG reel end to end — research the day's angle, write the avatar script, generate the avatar video, render it with captions, and prep the caption + guide link, held for my approval before posting

  Daily IG reel: produce and render