face-slap-drama-producer · öffentlicher Open Skill

face-slap-drama-producer

face-slap-drama-producer ist ein öffentlicher Open Skill auf Digen. Öffne ihn, um zu sehen, was dieser Workflow erstellt.

@jeneek3337

ShortDrama

13 2 10

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/2173956/2026-09-25/8b4cj24f.png

# Face-Slap Drama Producer

Creates a continuous, multi-episode "underdog-reversal" (face-slap) drama series: an underestimated protagonist who is belittled by others, gradually reveals hidden power through escalating "face-slap" moments, and ends in a full status reversal. Produces the full pipeline: locked script → character & scene reference images → many ~10s video clips per episode → auto-mixed episode cuts.

## Reference files
- `references/workflow.md`: full step-by-step production workflow (script → refs → clips → mix). Read it before starting a new series.
- `references/template.md`: the story-structure template (protagonist's hidden power, reason he's looked down on, antagonists, tone, setting, female lead, ending). Read it when the user gives a one-liner or partial idea, to fill in the blanks.

## When to use
- User wants a face-slap / underdog-reversal drama series (underestimated → rising up → complete reversal), from a one-liner, a filled template, or a partial idea.
- User wants to revise the script, swap a character image, redo one clip, or continue to the next episode of an existing series.

## Steps
1. First call `write_todos` to lay out the pipeline steps, and update it as you progress.
2. Gather the story parameters (hidden power, reason for being looked down on, antagonists, tone, setting, female lead, ending). If the user gave only a one-liner, fill in the blanks per `references/template.md` and confirm the outline before generating.
3. Lock the script: write the full continuous series as a document via `create_document(content, name="<series> - Script", entity_type="scripts")`. Break the story into episodes (no fixed limit — as many as needed for smooth flow), and break each episode into many short clips (typically 8–15, each ~10 seconds) so the flow is smooth. **Write rich, annoying dialogue into the script**: the antagonists' insults, belittling, and humiliation lines must be vivid and grating so the viewer genuinely feels annoyed and invested in the protagonist's revenge. Dialogue is carried in the script text only (no voiceover); the video clips are visual.
4. Generate character reference images: `generate_image(model='t2i', orientation='portrait', entity_type='characters', name='<character>')` for the protagonist, female lead, and each key antagonist. Keep each character's look consistent and reuse the same returned asset_id as a reference in all later steps.
5. Generate scene/location reference images: `generate_image(model='t2i', orientation='portrait', entity_type='locations', name='<scene>')` for each recurring location.
6. For each clip, generate the video directly from the character and scene references: `generate_video(model='ref2v', input_urls=[<character ref(s)>, <scene ref>], orientation='portrait', duration=10, entity_type='storyboard_videos', name='<episode> - clip <n>')`. Each clip is a ~10-second video. **Write the motion/action prompt to include a transition in (how the shot opens, leading from the previous clip) and a transition out (how it ends, leading into the next clip)** so the flow stays smooth — never a sudden change of location. **Embed the clip's dialogue into the video prompt, written in the same language as the script**, so the generated video carries that language (characters speaking it and/or on-screen dialogue text). **Do not use TTS/voiceover.** ref2v preserves character identity from the reference images.
7. **Auto-mix after each episode**: once all of an episode's clips are generated, automatically mix them into one episode video with `mix_clips(clips=[...in story order...], orientation='portrait', entity_type='final_videos', name='<episode>')`. Do this for every episode — each episode always yields one final mixed video.
8. Summarize the episode in natural language; call `set_guidance` to suggest next actions (next episode, redo a clip, swap a character, revise script).

## Notes
- **Smooth transitions are mandatory**: clips must flow into each other — never cut ab