Skip to content
Conduit

skill

shuodedaoli-skill

jiayuqi7813/shuodedaoli-skillcommunitylistedunreviewed
Use skill

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/jiayuqi7813/shuodedaoli-skill itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 26, 2026

dialogue-shorts

dialogue-shorts is a small toolkit for generating vertical dialogue videos with:

  • a 9:16 gameplay or background clip
  • 1-5 talking character portraits
  • colored ASS subtitles
  • optional Fish Audio TTS with local per-line audio cache reuse

It is designed to work both as a standalone Python tool and as a Codex/Cursor skill.

Features

  • Generate preview videos without TTS
  • Generate full voiced videos with subtitle timing driven by audio duration
  • Reuse cached local audio when only layout, subtitles, portraits, or background change
  • Auto-wrap long subtitles to stay inside the mobile-safe area
  • Support character portrait generation through image tools or your own assets

Project Structure

dialogue-shorts/
├── SKILL.md
├── reference.md
├── assets/
│   ├── backgrounds/
│   └── characters/
├── examples/
├── output/
├── scripts/
└── video_framework/

Requirements

  • Python 3.10+
  • ffmpeg
  • yt-dlp

Install dependencies:

pip install -r requirements.txt

Environment

Create a local .env file in the repo root when you want voiced output:

FISH_API_KEY=your_key

Quick Start

Preview only:

python3 scripts/generate_preview.py examples/sample_script.yaml -o output/preview.mp4

Full voiced render:

python3 scripts/generate_full.py examples/sample_script.yaml -o output/full.mp4

Audio Cache

Full renders cache each line of TTS audio under:

output/audio_cache/<script_stem>/

That directory stores:

  • line_000.mp3, line_001.mp3, ...
  • one JSON metadata file per line
  • manifest.json

If the line text, speaker, and voice_id stay the same, the next full render reuses local audio instead of calling Fish TTS again.

Background Assets

The repo does not bundle a large sample background video by default. Put your own source clip under assets/backgrounds/, or download one with yt-dlp as described in reference.md.

Skill Usage

The Codex/Cursor skill instructions live in SKILL.md.

License

MIT

↑↓ navigate openesc close