Micro Horror Drama (50s) · compétence Open Skill

Micro Horror Drama (50s)

Micro Horror Drama (50s) est une Open Skill publique sur Digen. Ouvrez-la pour voir ce que ce workflow crée.

@moviehub24542

ShortDrama

7 0 2

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4068827/2026-10-16/2b4p2w2r.jpg

---

allowed-tools:
  - generate_image
  - generate_video
  - generate_tts
  - generate_music
  - mix_clips
  - create_document
  - update_document
  - get_document
  - recognize_image
  - list_models
---

# Micro Horror Drama (50 seconds)

Turn a one-liner horror idea into a complete ~50-second vertical (9:16) micro drama: script → character sheet → storyboard images → video clips → voiceover + music → title cover → final cut with burned-in subtitles. All output is original and copyright-safe.

## When to use
- User gives a horror one-liner / logline / idea and wants a short-form horror video (~50s).
- User wants to revise one scene, redo a clip, restyle a character, or continue to the next micro episode.
- User uploads a character still / reference image to base the lead on.

## Reference files
- `references/horror-prompt-templates.md`: Reusable English prompt templates for every generation stage (character, keyframe, clip, voiceover, music, title cover) plus a mood/palette bank. Read it before writing any generation prompt, and fill the `<placeholders>` with the current story's details.

## Steps

### 0. Plan
First call `write_todos` to list the stages below and update it as you progress. Keep at most one stage `in_progress` at a time.

### 1. Script (create_document)
Write a tight ~50-second horror micro-drama script. Structure it as **3 beats** (setup → escalation → twist/cliffhanger), roughly 15s / 20s / 15s. Keep dialogue minimal and punchy; rely on visual dread and a final twist. Include per-beat: scene description, character action, one short line of dialogue (if any), and the intended mood/sound.
- Call `create_document(content=<markdown script>, name="Micro Horror Script", entity_type="scripts", content_type="text/markdown")`.
- Show the user the script and ask for approval before generating visuals. If they want changes, `update_document(asset_id=<script id>, content=<revised>)`.

### 2. Lead character sheet (generate_image)
Design the protagonist (and any key monster/antagonist) as a consistent reference image. Read `references/horror-prompt-templates.md` for the character template.
- Use `generate_image(model='t2i', prompt=<character template filled in>, orientation='portrait', resolution='720P', name="Lead Character", entity_type="characters")`.
- If the user uploaded a character still, use `generate_image(model='i2i', input_urls=[<uploaded asset_id>], prompt=<restyle/keep>, orientation='portrait', name="Lead Character", entity_type="characters")` instead.
- Keep this image as the reference for all later clips to preserve identity.

### 3. Storyboard images (generate_image)
For each of the 3 beats, generate a keyframe image using the character reference for consistency. Use the keyframe template from the reference file.
- Use `generate_image(model='i2i', input_urls=[<character asset_id>], prompt=<keyframe template filled in>, orientation='portrait', resolution='720P', name="Beat N Keyframe", entity_type="storyboard_images")` for each beat.
- 3 keyframes total (one per beat). Keep them visually distinct but tonally consistent (same palette, same character).

### 4. Video clips (generate_video)
Turn each keyframe into a short motion clip (~5–8s each) using reference-to-video for identity consistency. Use the clip template from the reference file.
- Use `generate_video(model='ref2v', input_urls=[<keyframe asset_id>], prompt=<clip template filled in>, orientation='portrait', resolution='720P', duration=<5-8>, name="Beat N Clip", entity_type="storyboard_videos")` for each beat.
- Aim for 3 clips totaling ~50s (e.g. 15s / 20s / 15s). If a single clip can't reach the target length, generate multiple clips per beat and stitch them later.

### 5. Voiceover (generate_tts)
Generate a short narration/voiceover line for the drama (optional but recommended for horror tone). Pick a `voice_instruction` from the reference file.
- Use `generate_tts(model='tts.design', text=<narration or dialogue line>, voice_instruction=<English