Cartoon · skill pública Open Skill

Cartoon

Cartoon é uma Open Skill pública no Digen. Abra para ver o que este fluxo cria.

@customercareismail.s

Anime

32 4 31

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3188428/2026-08-30/7ygssdxb.png

# Cartoon Video Generator

A 4-stage pipeline: Characters → Story → Scenes → Video.

**Goal:** Generate a complete cartoon video using Higgsfield MCP — create characters as
Reference Elements, auto-write a scene-by-scene story, and fire `generate_video` for
every scene automatically.

**UX is button-driven.** Every routing question uses `ask_user_input_v0` with concrete
options. Never ask open-ended typed questions for navigation or confirmations.

**Keep internals silent.** Never mention tool names, UUIDs, MCP calls, or media_upload
steps to the user. Narrate only in plain, friendly language.

---

## STAGE 0 — Onboarding (Single Round-Trip)

Ask ALL onboarding questions in ONE `ask_user_input_v0` call before doing anything else.

**Q1 — Characters**
- "I'll describe my characters"
- "Use a known theme/story (I'll name it)"
- "No idea — Claude picks for me"

**Q2 — Story tone**
- "Fun & silly (kids style)"
- "Adventure & action"
- "Heartwarming & emotional"
- "Funny & sarcastic"

**Q3 — Number of scenes**
- "Short (3–5 scenes)"
- "Medium (6–10 scenes)"
- "Long (10–15 scenes)"

**Q4 — Cartoon visual style**
- "Pixar-style 3D"
- "2D flat animation"
- "Anime"
- "Storybook illustration"

If the user already gave characters/theme in their opening message, skip Q1 and infer
the characters yourself. Still ask Q2–Q4.

---

## STAGE 1 — Character Creation (Higgsfield Reference Elements)

### Step 0 — Probe available presets *(internal — silent)*

Before generating any image, call `Higgsfield:presets_show` to get available image
generation presets. Cache the results. For character images, prefer:
- `nano_banana_2` (Nano Banana Pro) — best for character consistency
- `seedream_v4_5` (Seedream 4.5) — fallback

### Path A: User Describes Characters

1. Confirm back a clean character card showing: Name, Appearance, Personality
2. For each character, silently:
   a. Call `generate_image` with model `nano_banana_2` and prompt:
      `"[Name], [appearance], [personality hint], cartoon character, full body,
      solid plain white background, facing forward, [art style] style, clean design"`
   b. Wait for the image job to complete
   c. Call `show_reference_elements` (action=`create`) passing the completed image job
      id and url as media reference — this registers the character as a reusable element
   d. Cache: `{ name, element_id, element_name }` for each character
3. Show a character sheet to the user — name, thumbnail (via `job_display`), brief bio
4. Ask: "Characters are ready! Build the story now?"
   - "Yes, let's go!" / "Tweak a character" / "Add another character"

### Path B: Claude Invents Characters

1. Based on chosen tone, auto-generate 2–3 characters with names, appearance,
   personality. Examples for each tone:
   - **Fun & silly:** a tiny red panda with a rocket backpack; a round blue robot with
     googly eyes; an old wizard frog who sneezes spells
   - **Adventure & action:** a teenage girl with glowing gauntlets; her sarcastic
     talking sword; a shadowy villain who collects stolen laughs
   - **Heartwarming:** a lonely lighthouse keeper; a stray cat who can read maps;
     a message in a bottle that writes back
   - **Funny & sarcastic:** an overworked office dog; his even-more-overworked cat
     manager; the office plant who judges everyone silently
2. Show the proposed lineup and ask:
   - "Love them — use these!" / "Swap one out" / "Generate completely new ones"
3. Once approved, proceed same as Path A — generate images + create Reference Elements

---

## STAGE 2 — Story Generation (Auto)

Claude auto-generates a full scene-by-scene story. Format:

```
TITLE: [Catchy story title]
LOGLINE: [1-sentence hook]

SCENE 1 — [Scene title]
  Setting: [Location + time of day]
  Characters: [Who appears]
  Action: [What happens — 2–3 sentences]
  Dialogue: [1–2 key lines]
  Mood: [e.g. playful, tense, heartwarming]
  Video prompt: [Ready-to-use Higgsfield prompt — see Stage 3 for format]

SCENE 2 — ...
[Continue for all