arca-marketing-video 2.5.0 → 2.6.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "arca-marketing-video",
-  "version": "2.5.0",
+  "version": "2.6.0",
   "description": "Brand-driven short-form marketing content kit: four installable Claude Code skills (carousel generator, storyboard prompt, video prompt, shorts editor) plus a shared brand profile and assets. Run `npx arca-marketing-video` to install them into .claude/skills.",
   "keywords": [
     "skill",

package/skills/storyboard-prompt/SKILL.md

@@ -700,7 +700,11 @@ The grid:
   edge-to-edge or with a thin uniform gutter, so every frame crops out cleanly on its own.
 - Frames in beat order (frame 1 = top-left). Each frame is a different moment (the 9 beats below).
 - Render the frames in the chosen VIDEO TYPE's look (UGC phone-shot by default — raw, imperfect,
-  human per Phases 6 and 9; cinematic / animation / product film per the declared type).
+  human per Phases 6 and 9; cinematic / animation / product film per the declared type). For UGC, the
+  frames must read like stills grabbed from a phone video, NOT stock/AI photos — apply Phases 6 & 9
+  (real skin texture, non-model casting, mixed practical light, off-center handheld framing, lived-in
+  clutter); avoid even studio light, model-perfect symmetric faces, plastic skin, and staged centering.
+  (`video-prompt` reuses these frames as start frames, so the realism has to start here.)
 
 The frames must contain NO:
 - panel numbers, labels, titles, or letters

package/skills/video-prompt/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: video-prompt
-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.
+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, forces phone-camera realism on the start frames (so they don't look like stock/AI photos), generates multi-angle coverage so the edit can cut fast like real UGC, 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
@@ -115,6 +115,29 @@ Step A — produce a start frame per shot (image model, `imageAI` node):
   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.
+
+PHONE-CAMERA REALISM FOR START FRAMES (UGC types — non-negotiable)
+The #1 failure mode is a start frame that looks like a polished STOCK PHOTO or AI render (even soft
+studio light, model-perfect symmetric faces, plastic skin, centered staged composition, sterile clean
+room). The video model inherits the start frame's look, so a stock-looking frame produces a stock-
+looking clip. Fix it at the frame.
+- Apply the SAME proven realism rules this skill already specifies for the footage — PHONE-SHOT VISUAL
+  STYLE, CHARACTER REALISM, FRAMING, LIGHTING, and DEPTH OF FIELD (all below) — to the START FRAME
+  too. The frame is a paused still from that same casual phone video, NOT a separate photo shoot. In
+  particular reuse: natural skin texture / pores / asymmetry, non-model casting, candid mid-action
+  expression (CHARACTER REALISM); mixed practical light, uneven exposure, blown-out window, mild
+  noise (LIGHTING); off-center imperfect handheld framing (FRAMING); normal phone depth of field, no
+  creamy bokeh (DEPTH OF FIELD); real lived-in clutter (ENVIRONMENT).
+- Add explicit phone-camera tokens to the image prompt: "shot on an iPhone, candid vertical phone
+  snapshot, amateur, mild phone HDR, slight grain, imperfect framing."
+- Always include a negative prompt: stock photo, AI render, 3D render, glossy, airbrushed, retouched,
+  model, supermodel, perfect skin, symmetrical, studio lighting, softbox, beauty lighting, creamy
+  bokeh, shallow depth of field, hyperdetailed, 8k, cinematic, magazine, advertisement, clean, sterile,
+  staged.
+- Applies whether you upscale a photo (path A) or design a fresh frame (path B). If the source/look is
+  already stock-ish (the common failure), explicitly degrade it toward phone realism — grain, uneven
+  light, candid expression, off-center framing — BEFORE generating the clip. (NON-UGC types: follow
+  that type's craft instead — this phone-realism bar is for UGC.)
 - **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
@@ -343,6 +366,33 @@ Avoid:
 Every shot should answer:
 “Why would someone keep watching right now?”
 
+FAST CUTS & MULTI-ANGLE COVERAGE (the UGC editing engine)
+
+A single continuous clip per beat reads as a slow AI ad. Real engaging TikTok/Reels UGC is CUT FAST
+and constantly SWITCHES ANGLE and shot size. This operationalizes the "new beat every 2–4s" rule and
+the CAMERA RULES + TRANSITIONS sections below for GENERATION — and it overrides the one-shot-per-panel
+default in SHOT-BY-SHOT EXECUTION: a storyboard panel is a BEAT, usually several shots, not one clip.
+You cannot cut between angles you never generated.
+
+Rules:
+- Cut roughly every 1–2.5s. No talking shot holds longer than ~3s without a cut, angle change, or
+  reframe. The whole video should feel like many short shots, not a few long ones.
+- Cover each beat from MULTIPLE angles / shot sizes, then cut between them. Generate 2–3 variations per
+  beat drawn from the FRAMING + CAMERA move set: wide/establishing → medium close-up → tight close-up →
+  over-the-shoulder → desk-level POV → reaction insert. Vary framing on EVERY cut (never two same-size
+  shots back to back).
+- Use the video model's multishot where available (Kling V3 multiShot up to 6 shots) to get angle
+  changes inside one generation; otherwise generate several short clips per beat with different framings
+  and assemble them.
+- Intercut B-roll / insert shots between talking shots — hands on the keyboard, the screen, the paper
+  packet, a face reaction, an object. These inserts let you cut on every sentence and hide jumps.
+- Punctuate with the proven TRANSITIONS vocabulary (jump cuts, reaction cuts, quick whip-pans, match
+  cuts, camera repositioning) and snap zooms / handheld punch-ins — never slow dissolves or cinematic moves.
+- Keep continuity locked across the angle changes (same person/face, wardrobe, set, lighting) via the
+  character profile — fast cuts must not become a different person.
+- The actual cutting/assembly is finished in the editor (`shorts-editor` skill): generate enough angle
+  coverage here so that stage can cut fast. Hand it one long clip per beat and it cannot.
+
 PHONE-SHOT VISUAL STYLE
 
 SCOPE: this section (and the CAMERA / FRAMING / LIGHTING / DEPTH OF FIELD / CHARACTER REALISM /
@@ -811,7 +861,8 @@ Avoid:
 
 SHOT-BY-SHOT EXECUTION
 
-For each storyboard panel:
+Each storyboard panel is a BEAT — realize it as several short shots from different angles, not one
+held clip (see FAST CUTS & MULTI-ANGLE COVERAGE). For each panel:
 
 - preserve the main action
 - preserve the emotional beat