SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillscreate-narrated-comic-strip-video

create-narrated-comic-strip-video

Design & creation

View repository

Updated 10-15 21:54

About this skill

create-narrated-comic-strip-video

Originally validated in content-goose at internal/gooseworks-ads/video-05-tom-and-jerry/. Adapt paths to your project.

Purpose

Encode the style and recipe for a specific video format that has worked for Gooseworks ads. This molecule does not invent its own pipeline — it constrains and configures the video-orchestrator/orchestrator skill to produce videos in this exact aesthetic.

If you want to produce ANY video ad, use video-orchestrator/orchestrator directly with your own creative brief. If you want to produce a video that looks and feels like the "Tom & Jerry meets felt mascot" pattern, invoke this molecule.

When to use

Use this molecule when ALL of these are true:

  • The ad is for a brand with a distinct physical/photographic mascot (felt toy, plushie, sculpture).
  • The story benefits from comedic exaggeration — chase, before/after, big reveal.
  • You want a deadpan-narrator-driven arc rather than a UGC talking head.
  • The product is something abstract or service-shaped (software, agency replacement, AI agent) that's hard to visualize as a real-world object.
  • The ICP responds to comedic relatability over slick UI demos.

Skip this molecule for: serious B2B trust pitches, ads that need product UI on screen, founder-led talking heads, real-product-photo ads (use the product-video molecule instead).

Style fingerprint

These constraints define the format. They are non-negotiable for this molecule — change one and you're making a different kind of video.

Visual

  • 2D cartoon characters (humans or animals), classic Looney Tunes / Cuphead-adjacent aesthetic. Hand-drawn ink linework, flat colors with subtle paper texture, exaggerated cartoon expressions, squash-and-stretch motion.
  • Photographic mascot stays photographic — never cartoonified. The clash between cartoon scenes and the real mascot photo IS the joke.
  • NO SCREENS. No dashboards, laptops, phones, message threads, UI overlays anywhere except the end card. Every "product effect" is rendered as a physical cartoon object: easels, billboards, magic poofs, scrolls, balance scales, money sacks, falling-leaf bills.
  • Backgrounds alternate between sunny cartoon park (sky + grass + sun) and pale-paper "studio" backdrops.

Audio

  • Deadpan narrator as the connective voice. Pixar-short narrator energy, never hype-announcer.
  • Cartoon caper music bed under the cartoon scenes — pizzicato strings, walking bass, light woodwinds. Pink Panther-meets-Curb-Your-Enthusiasm-lite.
  • Music drop to silence at the mascot hero shot. Single sub-hum on cut-in, then quiet.
  • Foley-only SFX. Slide whistles, paper rips, easel thunks, magic poofs, jaw-drop boings, scale slams, ka-chings, tumbleweeds. Zero UI / keyboard / phone / message-chime SFX.
  • Optional character voices for moments where a cartoon character speaks directly (whisper line, a one-line response from the mascot). These come from VOICES.md, distinct from the narrator.

Structure

  • Length: 25–45 seconds. Sweet spot is 30–35.
  • 10-scene template (adjust scene count to match arc):
    1. Hook — protagonist chases what they want, fails
    2. Setup — protagonist's frustration deepens
    3. Counter-protagonist — someone else has it easy
    4. The disparity — quantified visually (scale, money piles)
    5. Burst-in — protagonist demands the secret
    6. The ask — counter-protagonist asks the mascot
    7. Hero shot — mascot fills frame, music drops, optional one-line response
    8. The result — three poofs of magic, payoff visible
    9. Whisper line — protagonist breaks fourth wall, quotable line
    10. End card — pale gray, rounded mascot tile, headline, blue pill CTA
  • The whisper line at scene 9 is the quotable. Casting a separate character voice lifts it. Don't oversell.
  • End card snaps back to brand — strip cartoon energy, return to launch-video minimalism.

Inputs

  • A core concept paragraph that fits the style fingerprint.
  • Photographic mascot asset (e.g. <brand>/<mascot>.jpeg).
  • <brand>/brand_aesthetic.md for the end-card minimalism reference.
  • Output folder (e.g. <brand>-ads/video-NN-<slug>/).
  • Optional overrides: scene count, character names, tonal axis (deadpan vs absurd), end-card copy.

Workflow

This molecule is a thin wrapper on the orchestrator pipeline. Its job is to seed each step with the right style preferences.

  1. Pre-set the style locks. Before calling video-orchestrator/orchestrator, write a <video_folder>/.style-preferences.md with:

    • Cartoon style descriptor block (the exact phrase used in keyframe prompts)
    • Approved voices: narrator (e.g. Jon), one or two character voices (e.g. Tom for whisper, Goose for hero response)
    • Music cues: caper bed + brand-anchor bed for end card
    • SFX hard rule: foley only, zero UI sounds
    • End card spec: pale gray bg, rounded mascot tile, two-line headline, blue pill CTA below

  2. Bias brainstorm. When invoked through this molecule, the brainstorm step should already know the structural beats. Push back hard on any beat that requires a screen.

  3. Bias create-design-brief. Force a "Visual world rules" section that calls out: ads = painted poster boards on cartoon easels; money = physical cartoon objects; mascot = photographic; no screens. The world-rule names "ads as easels" and "money as bills + sacks" are the canonical metaphors.

  4. Bias create-storyboard. Use the cartoon CSS/SVG system from gooseworks-ads/video-05-tom-and-jerry/storyboard.html as the storyboard scaffold.

  5. Bias create-clips.

    • Keyframes via Higgsfield nano_banana_2 with the cartoon style descriptor block.
    • Animation via Higgsfield veo3_1_lite (4s clips, 9:16, start_image-driven).
    • Hero shot is photographic mascot animated via Higgsfield grok_video (image-to-video) on the actual mascot photo, OR FFmpeg Ken Burns zoompan if the talking-head feel isn't wanted.
    • End card is a PIL composite with rounded mascot tile + blue pill CTA.
    • VO renders use ElevenLabs voices from VOICES.md. Whisper line uses a separate character voice with stability ~0.40, style ~0.55. Hero-response line uses the mascot voice with stability ~0.45, style ~0.50.

  6. Bias edit-video. Music cue A (caper) under scenes 1–6; music drops to silence at hero shot, returns at scene 8; cue B (launch-video bed) crossfades in for the end card. SFX foley library only — no UI sounds.

  7. Bias review-video. Add one extra reviewer check for this style: "Does the felt mascot feel intentional or like a video glitch?" If the cartoon-vs-photo clash reads as broken, the cure is usually a half-second film-burn / record-scratch transition before the hero shot, OR a faint cartoon outline halo around the mascot.

Output

  • Same folder layout as video-orchestrator/README.md describes, plus:
    • <video_folder>/.style-preferences.md (this molecule's contribution)

Quality Checks

  • Style fingerprint matches: cartoon scenes, photo mascot, no screens, deadpan narrator, caper bed → silence → caper → launch-bed.
  • Whisper line lands as a quotable.
  • Hero shot reads as intentional (not as a broken upload).
  • End card snaps back to brand minimalism — pale gray, rounded mascot tile, blue pill CTA.
  • All voices used appear in VOICES.md.
  • Total runtime in the 25–45 s range; ≥ 80% of frames are cartoon-energy (only the hero shot and end card break that).

Failure Modes

  • Mascot gets cartoonified somewhere — re-export from the original photo source.
  • A "screen" sneaks in (phone, laptop, computer dashboard) — molecule fails its own constraint, restart that scene.
  • Music doesn't drop at the hero shot — joke lands flat. Re-mix that segment specifically.
  • Whisper line is read with hype-announcer energy — character-voice settings need lower stability and higher style.
  • End card looks like a regular cartoon scene with extra text instead of a real brand snap-back — strip the cartoon backdrop, return to brand minimalism.

Reference run

<your-project>/video-05-tom-and-jerry/ is the canonical execution of this molecule:

  • 10 scenes, 42.5s total
  • Cartoon scenes 1–6, 8, 9 via nano_banana_2 → veo3_1_lite
  • Hero shot (scene 7) via Higgsfield grok_video on goose3.jpeg — talking head with narrator + Goose VO ("What do you need?")
  • Whisper line (scene 9) using Tom character voice
  • End card (scene 10) — pale gray, rounded felt-Goose tile, "Stop chasing winners. / Start Goosing." headline, blue gooseworks.ai pill CTA, no VO
  • Voices: Narrator Jon · Goose QzTKubutNn9TjrB7Xb2Q · Tom mdzEgLpu0FjTwYs5oot0

Composed Atoms

  • atoms/voiceover/create-voiceover-elevenlabs — narrator + optional character voices from VOICES.md
  • atoms/image-generation/create-image-nano-banana-fal — cartoon keyframes with the locked style descriptor
  • atoms/video-generation/create-video-veo3 — 4s i2v clips, 9:16, start_image-driven
  • atoms/assembly/stitch-videos-ffmpeg — concat + music cue sequencing + foley SFX layering
  • atoms/end-cards/create-end-card-from-reference — pale gray brand snap-back with rounded mascot tile + blue pill CTA
  • atoms/audio-editing/mix-master — caper bed → silence at hero shot → caper → launch-bed mix
  • atoms/captions/burn-in-captions — optional narrator captions

Decision Rules

  • Hard rule — NO screens, NO UI, NO phones, NO dashboards anywhere except the end card. Every "product effect" must be a physical cartoon object (easel, billboard, magic poof, scroll, balance scale, money sack).
  • Photographic mascot stays photographic — never cartoonified. The clash IS the joke.
  • Narrator is deadpan Pixar-short energy — never hype-announcer.
  • Music drops to silence at the hero shot (scene 7); single sub-hum on cut-in, then quiet.
  • Foley-only SFX (slide whistles, paper rips, easel thunks, magic poofs); zero UI / keyboard / phone / message-chime SFX.
  • End card snaps back to brand minimalism — strip cartoon energy, return to brand voice + blue pill CTA.

This molecule is a thin wrapper that configures a video orchestrator with the cartoon-comic style. After style preferences are written, the orchestrator handles all sequencing, gating, and looping. To produce another video in this style for a different brand, change the mascot asset, brand_aesthetic, voices, and concept; everything else in the recipe stays the same.