1.4K ·GitHub
What is this?
This is an AIGC skill pack that gives AI agents a full industrial pipeline for writing and directing viral Chinese short dramas: script structuring, character/prop asset locking, storyboarding, dialogue tone-checking, martial-arts/fantasy combat choreography, and ready-to-feed prompts for video-generation models like Seedance and MiniMax H3.
This is an AIGC skill pack that gives AI agents a full industrial pipeline for writing and directing viral Chinese short dramas: script structuring, character/prop asset locking, storyboarding, dialogue tone-checking, martial-arts/fantasy combat choreography, and ready-to-feed prompts for video-generation models like Seedance and MiniMax H3.Is this an app, or a rule set I load into an AI agent?It is not a standalone app. It is a SKILL.md rule file plus reference modules that you load into an AI agent platform, and the agent then runs the P0-P5 pipeline, asset ledgers, and prompt templates automatically.
It is not a standalone app. It is a SKILL.md rule file plus reference modules that you load into an AI agent platform, and the agent then runs the P0-P5 pipeline, asset ledgers, and prompt templates automatically.How do I start using it?Load short-drama-director/SKILL.md and its reference modules into one of the supported AI agent platforms (OpenClaw, WorkBuddy, Doubao Coze, or Dify), and the agent runs the Asset-First P0 to P5 production pipeline from there.
Load short-drama-director/SKILL.md and its reference modules into one of the supported AI agent platforms (OpenClaw, WorkBuddy, Doubao Coze, or Dify), and the agent runs the Asset-First P0 to P5 production pipeline from there.Which video-generation models does it target?It is built around Seedance 2.5 as the primary render backend, with deep prompt adaptation for 15-second clips, first-last frame pairs, Hit-Stop effects and native [SFX:] tags; the project description also lists MiniMax H3 as an alternate model format.
It is built around Seedance 2.5 as the primary render backend, with deep prompt adaptation for 15-second clips, first-last frame pairs, Hit-Stop effects and native [SFX:] tags; the project description also lists MiniMax H3 as an alternate model format.Does it guarantee my content passes platform moderation?No. It includes a safety translation dictionary meant to lower platform risk, but it explicitly does not promise a 100% pass rate, and the project itself is labeled a test/beta release.
No. It includes a safety translation dictionary meant to lower platform risk, but it explicitly does not promise a 100% pass rate, and the project itself is labeled a test/beta release.What is the 'Asset-First' pipeline?It locks character, prop, and scene reference images before any storyboarding happens, so every shot inherits consistent visuals from CHR/PRP/SCN ledgers instead of being improvised scene by scene.
It locks character, prop, and scene reference images before any storyboarding happens, so every shot inherits consistent visuals from CHR/PRP/SCN ledgers instead of being improvised scene by scene.Why does it have a martial-arts library?It is aimed at action-heavy wuxia/xianxia short dramas, so it bundles 23 martial-arts schools, 9 sword forms, and an R1/R2/R3 scoring system for 15-second fight shots to keep combat scenes from losing viewer completion.
It is aimed at action-heavy wuxia/xianxia short dramas, so it bundles 23 martial-arts schools, 9 sword forms, and an R1/R2/R3 scoring system for 15-second fight shots to keep combat scenes from losing viewer completion.Why does it forbid static scenes?Current AI video models cannot convincingly 'act' while standing still, so the V6.5 update added an iron rule that emotional beats and energy build-up must happen through motion, never a static pose.
Current AI video models cannot convincingly 'act' while standing still, so the V6.5 update added an iron rule that emotional beats and energy build-up must happen through motion, never a static pose.What does the quality-check stage catch?Before a prompt is fed to the render model, it runs independent P0/P1/P2 gates plus a manuscript-level machine check (C1-C20 / H1-H12) backed by a 41-rule library and an offline storyboard dashboard.
Before a prompt is fed to the render model, it runs independent P0/P1/P2 gates plus a manuscript-level machine check (C1-C20 / H1-H12) backed by a 41-rule library and an offline storyboard dashboard.What are similar projects?shuohao-skills, seedance-2-5-video-director, reelmimic, autoclip, film-studio-skills, lanshu-create-ai-presenter-video, short-drama-production, h3-storyboard-skill
shuohao-skills, seedance-2-5-video-director, reelmimic, autoclip, film-studio-skills, lanshu-create-ai-presenter-video, short-drama-production, h3-storyboard-skillWhich topics is it in?dify, storyboarding, ai-prompts, creative-writing, minimax-h3, douyin, ai-video, script, video-editing, content-creation, workflow-automation, video-generation
dify, storyboarding, ai-prompts, creative-writing, minimax-h3, douyin, ai-video, script, video-editing, content-creation, workflow-automation, video-generation