Kitten to Giant Lion Viral Video · Open Skill 공개 스킬
Kitten to Giant Lion Viral Video
Kitten to Giant Lion Viral Video은 Digen의 공개 Open Skill입니다. 열어 보면 이 워크플로가 무엇을 만드는지 알 수 있습니다.
Roleplay
1 0 4
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3766200/2026-10-20/pnksyqrg.jpg
--- allowed-tools: - generate_image - generate_video - generate_tts - generate_music - mix_clips - caption_video - create_document - recognize_image --- # Kitten to Giant Lion — Viral Transformation Video A "cute-to-mighty" transformation short: a tiny kitten in a dangerous, fire-lit world grows into a giant lion and takes command of it — by default it lands as a **bold half-cat half-lion hybrid**, one side tiny kitten, one side giant lion, split down the middle. Default output is a cinematic landscape 16:9 cut (hook in the first second, escalating scale, tension build, payoff); switch to `portrait` 9:16 only when the user explicitly asks for a TikTok/Reels/Shorts vertical. ## When to use - The user describes a "tiny animal becomes a giant" transformation video (kitten → lion especially), often in a burning/ash/fire setting. - The user wants a finished cut, or wants one stage reworked (new beats, restyled shot, new voiceover, new music, new captions, new cover). ## Reference files - `references/beat-sheet.md` — the 8-beat narrative template with tension beats, the hybrid-form prompt rules, cover-art directions, and the re-style/re-do recipes. Read it before writing the beats or prompts. ## Steps 1. **Scope it.** Ask only if truly missing: aspect ratio (default `landscape` 16:9 cinematic; use `portrait` 9:16 only if the user asks for a vertical short), duration target (default 30–40s total), tone (default: awe-inspiring, dramatic, cinematic), final form (default: bold half-cat half-lion hybrid; can also be a full lion or a full cat), and whether a voiceover is wanted (default: yes, short punchy narration). Otherwise proceed with defaults and state them in your reply. 2. **Plan.** Use `write_todos` to list the stages below and update the whole list as you progress. Write the beat sheet with `create_document` (content = Markdown beats; `name="Kitten to Lion — Beat Sheet"`, `entity_type="scripts"`) so the user can review/tweak the narrative. 3. **Storyboard frames.** For each beat generate one key frame with `generate_image`, `model='t2i'`, `orientation='landscape'`, `resolution='1080P'`, `name="SB <n> — <beat title>"`, `entity_type="storyboard_images"`. Keep the kitten's markings and the lion's mane coloring identical across beats (repeat the exact description of the fur pattern in every prompt) so the transformation reads as one character. Add `input_urls` of the previous frame with `model='i2i'` only when the user asks for a restyle of a specific frame. 4. **Animate each beat.** `generate_video` with `model='ref2v'` (reference mode: all `input_urls` are reference images, never start/end frames) and the storyboard image of that beat as the reference. Pass `orientation='landscape'` and `duration` (4–9s per beat, sum to the target length; shorten the tension beats for a rising cut rate). `name="SHOT <n> — <beat title>"`, `entity_type="shot_videos"`. Prompt for wide cinematic camera motion (sweeping push-in, low angle orbit, slow-motion crane pull-back) and keep the fire/jungle lighting consistent. Alternative channels: `model='i2v'` for a straight single-image animation, `model='t2v'` (explicit) only if no reference image exists. 5. **Voiceover.** `generate_tts` with `model='tts.design'` and a `voice_instruction` in English (e.g. "cinematic trailer voice, deep and warm, breathless awe, slow deliberate pacing"). `name="Voiceover"`, `entity_type="voiceovers"`. Keep the script under ~60 words, giving each tension beat its own short line. If the user supplied a reference voice, use `model='tts.clone'` with `reference_audio_url` instead. 6. **Music.** `generate_music` with `model='music'`, prompt for a cinematic tribal-percussion / orchestral hybrid that builds tension, drops to silence before the transformation, then swells. `name="Score"`, `entity_type="music"`. 7. **Assemble.** `mix_clips` with the shot clips in order, `videos` (one entry per shot, with `duration` matching that shot) and `audios` for voi