arca-marketing-video 2.3.0 → 2.4.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "arca-marketing-video",
-  "version": "2.3.0",
+  "version": "2.4.0",
   "description": "Brand-driven short-form marketing content kit: four individual agent skills (carousel generator, storyboard prompt, video prompt, shorts editor) plus a shared brand profile and assets.",
   "keywords": [
     "skill",

package/skills/video-prompt/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: video-prompt
-description: Use when turning a 3×3 storyboard into a finished vertical short-form video of any type (UGC, cinematic / movie-trailer, animation, product film, etc.) via the Wyren MCP — picking image/video models and resolutions, optimizing the TikTok first 5 seconds, keeping character faces consistent across shots, and generating ≤15s clips. Triggers on "make the video from this storyboard", "generate the short", "render the clips". Part of the arca-marketing-video kit.
+description: Use when turning a storyboard into a finished vertical short-form video of any type (UGC, cinematic / movie-trailer, animation, product film, etc.). Handles both photographic storyboards (clean and use as start frames) and schematic / annotated plans (do NOT upscale 1:1 — rebuild patterned to them using Wyren clips for live footage + HyperFrames for on-screen graphics). Drives the Wyren MCP — picks image/video models and resolutions, optimizes the TikTok first 5 seconds, and keeps character faces consistent across shots. Triggers on "make the video from this storyboard", "generate the short", "render the clips". Part of the arca-marketing-video kit.
 ---
 
 # Video Prompt
@@ -14,7 +14,7 @@ Create a realistic vertical short-form video from the provided 3x3 storyboard re
 
 INTAKE — ASK FIRST:
 Before generating, ask the user for any of these that are not already provided, then wait for their answers:
-1. STORYBOARD REFERENCE — the 3x3 storyboard image (and shot notes if available).
+1. STORYBOARD REFERENCE — the storyboard image (and shot notes if available). Note whether it is PHOTOGRAPHIC frames or a SCHEMATIC / annotated plan (panel numbers, notes boxes, mock UI) — see STORYBOARD INTERPRETATION; it changes how the video is built.
 2. VIDEO TYPE / STYLE — UGC / phone-shot (default), cinematic / movie-trailer, animation, motion-graphics, product film, skit, etc. Should match the storyboard's declared type. Drives the LOOK and the first-5-seconds cold open. If UGC, the phone-shot styling below applies; if not, follow that type's craft.
 3. BRAND PROFILE — the attached or pasted ../../brand.md (brand name, logo rules, persona, colors).
 4. TARGET MARKET / AUDIENCE — who this is for.
@@ -26,7 +26,12 @@ Before generating, ask the user for any of these that are not already provided,
 10. NATIVE AUDIO — whether the video model should synthesize dialogue/sound (only some models support it). Default: on if the chosen model supports `sound`.
 Ask first. Only make smart, briefly stated assumptions for whatever is still missing, and state them briefly. If the user names a budget or "cheapest/fastest", pick the model tier accordingly and say which you picked.
 
-Use the storyboard as the primary visual anchor. Crop the frames, upscale and clean them using image AI and then use those as the first frames. You may generate clips separately video merge them (or merge them using Hyperframes), or you may also do multishot for Video AI decide automatically and smartly.
+Use the storyboard as the PLAN, not as footage to copy. First classify it (see STORYBOARD
+INTERPRETATION below): if the panels are photographic, clean them and use them as start frames; if
+they are schematic / annotated mockups (panel numbers, notes boxes, drawn phone bezels, mock UI), do
+NOT upscale or reproduce them 1:1 — rebuild the video patterned to them, using Wyren clips for the
+live footage and HyperFrames for the on-screen graphics. You may generate clips separately and merge
+them (Wyren or HyperFrames), or use video-model multishot — decide smartly.
 
 VIDEO TYPE SCOPE — read before applying the rules below.
 This skill makes ANY video type. The LOOK follows the chosen VIDEO TYPE; the TikTok retention rules
@@ -43,6 +48,42 @@ rules) apply to EVERY type.
   — render in the chosen type's craft (e.g. for a trailer: cinematic lighting, fast cutting, scale).
   Still keep the retention rules, continuity locks, brand/logo rules, and the first-5s cold open.
 
+STORYBOARD INTERPRETATION — PHOTOGRAPHIC vs SCHEMATIC (read before generating)
+
+Storyboards arrive in two forms. Detect which you have before touching Wyren.
+
+A) PHOTOGRAPHIC frames — each cell is a real / clean image of the actual scene (people, environment),
+   like the clean frame grid the `storyboard-prompt` skill now outputs.
+   → Crop each frame to 9:16 and clean / upscale it into a start frame (GENERATION SETTINGS Step A),
+     keeping the face, then use it as the videoAI startFrame. This is the 1:1-friendly path.
+
+B) SCHEMATIC / ANNOTATED storyboard — each panel is a rough PLAN: a drawn phone bezel, small webcam
+   tiles, mock UI cards, labels, sticky notes, captions, and a notes box (SCENE / TIME / SHOT / ANGLE
+   / MOVE / ACTION / DIALOGUE / AUDIO / TRANSITION / RETENTION / PHONE REALISM). The example
+   "AI Operator Interview" sheet is exactly this kind.
+   → Treat it strictly as a SHOT PLAN, never as pixels. Do NOT upscale it, do NOT reproduce panels
+     1:1, and never render the panel numbers, "Panel N" labels, notes boxes, drawn bezels, or the
+     panel's rough mock graphics as-is into the video. Instead REBUILD each beat:
+     - Read the panel's SCENE / ACTION / SHOT / ANGLE / MOVE / DIALOGUE / AUDIO / TRANSITION /
+       RETENTION as the brief for that shot.
+     - Generate the LIVE FOOTAGE with Wyren (people, faces, desk, reactions, real environment, camera
+       move) using the recurring character profile for face consistency and a fresh start frame
+       DESIGNED for that shot via image AI — not the schematic panel.
+     - Build the ON-SCREEN GRAPHICS with HyperFrames and composite them over the Wyren footage: UI
+       cards, data / deck mockups, split-desk labels, chips, checklists, route-map lines, captions,
+       progress arcs, a REC indicator, the logo — anything that is a graphic, not live action.
+     - Match each panel's layout and intent (what graphic sits where, what the person is doing), but
+       realize it as real footage + clean motion graphics in the chosen VIDEO TYPE's look.
+
+WHO MAKES WHAT (schematic path):
+- Wyren videoAI → live-action clips: people, faces, reactions, hands, environment, props, camera move.
+- Wyren imageAI → designed start frames and any photographic plates feeding the clips.
+- HyperFrames → all overlay graphics, captions, chips, UI / data mockups, transitions, logo, splash,
+  composited on top of the clips (the same graphics layer the `shorts-editor` skill uses).
+Keep diegetic vs overlay clear: a screen the actor really looks at can be a graphic comped onto the
+device; floating captions / chips are HyperFrames overlays. Never bake storyboard annotations into the
+video. If you are unsure which form the storyboard is, ask the user before generating.
+
 CORE OUTPUT
 
 Generate a finished vertical short-form video with:
@@ -67,8 +108,13 @@ options with `mcp__wyren__list_models` and `mcp__wyren__get_model_capabilities`
 the lists below are a current snapshot, not a contract, and per-model resolution/duration/startFrame
 support changes.
 
-Step A — clean storyboard frames into start frames (image model, `imageAI` node):
-- Crop each storyboard panel to 9:16, then upscale/clean it into a usable first frame.
+Step A — produce a start frame per shot (image model, `imageAI` node):
+- PHOTOGRAPHIC storyboard (path A): crop each panel to 9:16, then upscale/clean it into a usable
+  first frame.
+- SCHEMATIC storyboard (path B): do NOT upscale the panel. Use image AI to DESIGN a new start frame
+  for that shot from the character profile + the panel's brief (scene, action, framing), so the frame
+  is real-looking footage, not a redraw of the mockup. The panel's mock UI becomes a HyperFrames
+  overlay later, not part of this start frame.
 - **Keep the face.** When upscaling/cleaning with the image model, instruct it to PRESERVE the
   existing person's identity — same face, hair, age, build, and wardrobe — and only improve quality
   (sharpen, denoise, fix artifacts). Do not let it redraw or beautify into a different face. Use an
@@ -101,12 +147,18 @@ Step B — generate the clips (video model, `videoAI` node):
 
 Wyren execution flow (per the wyren skill's policy — load it before any `mcp__wyren__*` call):
 1. `list_models` + `get_model_capabilities` to lock the exact image/video model, resolution, mode, duration.
-2. `build_graph`: `imageInput` (storyboard frame, characters.png, logo.png) → `imageAI` (clean/upscale) →
-   `videoAI` (the chosen model/resolution/mode/duration). Use multi-shot or per-clip nodes per the split rule.
+2. `build_graph`: `imageInput` (start-frame source, characters.png, logo.png) → `imageAI` (path A: clean/
+   upscale the photo panel; path B: design a fresh start frame) → `videoAI` (the chosen model/resolution/
+   mode/duration). Use multi-shot or per-clip nodes per the split rule.
 3. `validate_workflow` — resolve warnings with the user.
 4. Estimate cost with `get_pricing` (chain mode) / `estimate_product_cost`; get the user's OK to spend.
 5. `run_workflow` (`userConfirmed: true`), then poll `get_workflow_run_status` every 5s until terminal.
-6. Pull results with `get_node_outputs`. Merge multi-clip output (Wyren or HyperFrames) into the final cut.
+6. Pull the clips with `get_node_outputs`.
+7. GRAPHICS PASS (HyperFrames) — build the on-screen overlay graphics the storyboard calls for (UI/data
+   cards, chips, captions, checklists, route maps, logo, transitions) and composite them over the Wyren
+   clips. For a SCHEMATIC storyboard this pass is required (the mock UI in the panels lives here, not in
+   the footage). Hand off to the `shorts-editor` skill, which owns this HyperFrames graphics + master step.
+8. Merge the clips + graphics into the final cut.
 
 RECURRING CHARACTER CONSISTENCY (multishot / multi-clip)