Green Snake Viral Story Director · публичный навык Open Skill

Green Snake Viral Story Director

Green Snake Viral Story Director — публичный Open Skill на Digen. Откройте, чтобы увидеть, что создаёт этот процесс.

@rajaajaiswarr456

ShortDrama

3 2 3

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4296757/2026-10-20/rf33md6e.jpg

---

allowed-tools:
  - generate_image
  - generate_video
  - generate_tts
  - mix_clips
  - create_document
---

# Green Snake Viral Story Director

Turns one of the fixed funny stories (or a new story the user describes) into a shootable
vertical comedy short: scene breakdown → image prompts → video prompts → generated clips →
final cut.

## Duration modes
- **10s (default)** = ONE 10-second clip, ONE scene, ONE location, one conflict.
  Beat structure: 0–1.5s hook, 1.5–6s escalation, 6–8s twist, 8–10s punchline.
  Only three things survive: hook line, twist, punchline line. Everything else is cut.
- **20s** = 2 clips of 10s (Google Flow AI style), Scene 1 = 0–10s, Scene 2 = 10–20s,
  with a continuity lock between them.

If the user says "just 10 sec", "short version", or "single clip", use 10s mode. If they
ask for a specific story, use it as the source; in 10s mode, compress the story to its
single strongest gag rather than trimming evenly.

## Cast flexibility (important)
The cast is user-configurable — the user can change the lead creature and add or swap animal
sidekicks for a specific story.
- Default lead = **bright-green gecko-snake hybrid** (gecko head, big eyes, toe pads, curled
  gecko tail, snake-like scaled body, black sleeveless gym vest, red headband).
- Default sidekick = the mouse (blue tracksuit, whistle, towel).
- If the user names other animals (dog, cat, frog, parrot, tortoise, rabbit…), give each one a
  distinct color palette, silhouette, one funny outfit or prop, and a locked identity. Extras
  are non-speaking.
- The cast that the user picks is LOCKED for that whole video: same face, colors, pattern,
  proportions, clothes and voice in every scene and clip. The green-color lock applies to the
  lead creature.

## Reference files
- `references/character-locks.md` — read before writing ANY prompt. Lead creature + sidekick identities, cast system, visual style, sound rules, dialogue rules, duration modes.
- `references/stories.md` — the fixed stories with full beat-by-beat dialogue, plus 10-second single-clip cuts. Read the story the user picks (e.g. "Story 3") plus any story listed there when the user says "all".

Read only the files you need for the current task, not all of them.

## Workflow (staged — do not skip ahead)

The user drives the stages with a story number (1–10) or a custom idea.

### Stage 1 — Scene breakdown (default, do this first)
Deliver ONLY the story + the scene breakdown (no image prompts yet):
- Full story summary with the hook, escalation, twist and punchline.
- 10s mode: one scene, beats timestamped inside 0–10s.
- 20s mode: Scene 1 = 0–10 sec, Scene 2 = 10–20 sec.
- Each beat carries: timecode, action, reaction, exact dialogue (Hindi, short, funny), who says it, sound effects, camera move.
- Name the speaking character explicitly per line.
- Apply the retention lock: a new visual/comedy beat every 2 seconds in 10s mode (every 2–3s in 20s).
- Then ask (via set_guidance) whether to proceed to IMAGE PROMPT.

### Stage 2 — Image prompts (only when the user asks)
Write a detailed image prompt per scene (English, ready to paste into an image model), containing:
- The full character lock text for every character in frame (species, identity, color, pattern, clothes).
- Environment, lighting, camera angle, 9:16 framing, 3D animated movie realism.
- Exact same face/body/pattern/clothes as the previous scene.
Ask before generating actual images unless the user says "generate".

### Stage 3 — Video prompts (only when the user asks)
Write the video prompt(s) — one per 10-second clip (10s mode = exactly one). Each contains ALL of:
continuous voice, continuous action, zero silence, exact dialogue with speaker name, lip-sync note, sound effects, camera movement, character consistency note, lead-color lock.

### Stage 4 — Generation (only when the user says generate/produce)
1. Character sheet images first: `generate_image(model='t2i', orientation='portrait', name='<Story> <Character> Charact