Syntha - AI Video Factory / AI 视频创作工厂

English | 简体中文

---

English

Syntha is a high-performance, automated AI video orchestration platform designed to streamline the production of short-form content. By integrating cutting-edge AI models for script writing, visual generation, and audio synthesis, Syntha transforms raw ideas into polished cinematic videos with a single click.

🌟 Key Features

• Multi-Scene Generation: Intelligent decomposition of scripts into sequential visual prompts for consistent storytelling.
• Dynamic Cinematic Synthesis: Advanced FFmpeg engine with smooth transitions (Zoompan) and automatic audio-visual alignment.
• Smart Audio Pipeline:
  • Edge-TTS: High-fidelity voice synthesis with multi-language support.
  • MiniMax BGM: AI-generated stylized background music.
  • Audio Ducking: Pro-level mixing that automatically lowers BGM volume during narration.
• Magic Brain: AI-powered script expansion and "Director Mode" for automated prompt engineering.
• Bilingual Subtitles: Automatic CJK-supported hard-coded subtitles with built-in translation.

🚀 Quick Start

1. Launch Services: Start the entire cluster (Frontend, LLM, TTS, Media) with a single command:

   ./dev.sh up

2. Access Dashboard: Open http://localhost:8102 in your browser.

🏗 Architecture

• Frontend: Remix (React) + Vite + Tailwind CSS.
• Orchestration: Node.js TaskCoordinator with real-time SSE progress tracking.
• AI Services: Python FastAPI microservices (LLM, TTS, BGM).
• Storage: S3-compatible (RustFS/MinIO) for immutable asset persistence.
• Database: SQLite (Prisma) with WAL mode enabled for high concurrency.

---

简体中文

Syntha 是一个高性能、全自动的 AI 视频编排平台，专为短视频内容的高效生产而设计。通过集成顶尖的脚本撰写、图像生成和音频合成 AI 模型，Syntha 能将原始创意一键转化为制作精良的电影感视频。

🌟 核心特性

• 多场景序列生成：智能拆解文案，生成连续的分镜头提示词，确保故事连贯性。
• 动态电影感合成：基于 FFmpeg 的高级合成引擎，支持平滑缩放转场（Zoompan）及音画自动对齐。
• 智能音频链路：
  • Edge-TTS：高保真语音合成，支持多国语言。
  • MiniMax BGM：AI 创作风格化背景音乐。
  • 自动闪避 (Ducking)：专业级混音，解说人声出现时 BGM 自动压低。
• 智写大脑 (Magic Brain)：AI 文案扩写与“导演模式”，实现全自动分镜提示词工程。
• 智能双语字幕：支持 CJK 字库的硬烧录字幕，集成一键中英翻译。

🚀 快速开始

1. 启动服务：通过一键脚本启动所有后端集群与前端：

   ./dev.sh up

2. 访问系统：在浏览器打开 http://localhost:8102。

🏗 架构设计

• 前端：Remix (React) + Vite + Tailwind CSS。
• 编排层：Node.js TaskCoordinator 配合 SSE 实现毫秒级进度同步。
• AI 服务：基于 Python FastAPI 的微服务集群 (LLM, TTS, BGM)。
• 存储：支持 S3 协议的对象存储 (RustFS/MinIO)，确保资产持久化。
• 数据库：SQLite (Prisma)，默认开启 WAL 模式以支持多服务高并发读写。

---

⚠️ Infrastructure Constraints / 基础架构约束

• Database: All services must connect to the absolute path of dev.db at the project root to prevent multi-instance database creation.
• Ports: TTS (8100), LLM (8101), Media (8104), Web (8102).
• Temp Files: All intermediate processing is handled in the system's temporary directory to ensure filesystem safety.

数据库说明：所有服务必须连接至项目根目录下的 dev.db 绝对路径，严禁在子目录下误创数据库。