Kids Jungle Safari cute animation · Open Skill 공개 스킬

Kids Jungle Safari cute animation

Kids cute animation short stories

@Vjagstya

Kids

11 3 3

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4136188/2026-09-23/m46qxt4d.jpg

# Coco & Kiki Jungle Safari

## When to use
- User wants a Coco & Kiki jungle safari clip or episode: the two kids (Coco, a 4-year-old boy, and Kiki, a 3-year-old girl) with their golden retriever puppy and gray kitten, spotting a lion and a tiger in a sunlit rainforest from their safari jeep.
- User wants to advance to the next episode, re-render, restyle, change orientation (landscape/portrait), or adjust the dialogue.

## Preset assets
- `safari_scene` — the reference scene image: Coco and Kiki in the tan safari jeep (Coco in a blue hoodie with a yellow star and tan safari vest holding the golden retriever puppy; Kiki in a tan pith helmet and pink flower shirt with the gray kitten peeking from her backpack), with a male lion resting on a mossy ledge and a tiger near a stream in a lush rainforest. Use this as the ref2v reference image.

## Steps
1. Clarify the brief (one-liner, episode number, or new idea). If no source text is provided, `search_web` (max 2 calls) for context.
2. Write the two gate blocks as visible markdown: `## Story plan` (Route / Window / Question / Cursor / Last Button / Deferred) and `## Beat budget` (Format / Story / Scene goal / Assigned / Deferred / Button). Do not render before both blocks exist.
3. Plan the world (sunlit tropical rainforest, whimsical storybook 3D animation, lush greens + golden dappled light), the entities (Coco, Kiki, the golden retriever puppy, the gray kitten, the lion, the tiger), and a 3–6 shot cut map.
4. Render the clip with `generate_video`:
   - `model='ref2v'`, `input_urls=[safari_scene]`
   - `duration=10` (or user override)
   - `orientation='landscape'` (default) or `'portrait'` per user
   - `resolution='720P'`
   - `entity_type='final_videos'`
5. Reply with the creative payoff (what the viewer reads changed) and one line on the production stack (ref2v · N references · N shots · duration · orientation). Do not paste asset IDs or URIs.

## ref2v prompt (six-part, English)
Write the prompt as six sections in order: `subject_definitions` (define `<Subject N>` tags from the reference), `summary` (task type + target), `retention_analysis` (fully_preserved per subject), `detailed_description` (style opening + `[Shot 1]..[Shot N]` with cut times summing to duration), `overall_soundscape` (jungle ambience: birdsong, rustling leaves, stream, breeze), `non_diegetic_music` (gentle whimsical orchestral score). Dialogue goes inside `<d>[English] ...</d>`.

## Notes
- Locked to 720P: always pass `resolution='720P'`; never pass `width`/`height`/`resolution_model`/`aspect_ratio`.
- Shots live in the ref2v prompt — do not generate storyboard images.
- Keep Coco and Kiki's look consistent with the reference across shots (messy brown hair, blue hoodie with yellow star, tan safari vest for Coco; tan pith helmet, pink flower shirt, pink backpack for Kiki).
- Multi-segment same episode (user asks >10s or per-shot): each segment `entity_type='storyboard_videos'`, then `mix_clips` → `entity_type='final_videos'`. Never mix clips from different episodes.
- Report failures honestly; never fabricate URLs.