Agions/dramaxis

🎬 What is Dramaxis?

Dramaxis (剧枢) is an open‑source desktop application that helps creators turn raw video footage into polished, AI‑generated short‑film or commentary videos. It combines a Rust‑native multi‑agent engine with a Tauri‑based desktop shell and a React 19 front‑end. The system orchestrates several specialized AI agents (director, visual analyst, screenwriter, voice artist, sound engineer, quality reviewer) to:

• analyse video frames and extract key moments,
• write a short, hook‑focused script,
• generate or clone a voice‑over,
• align audio‑video tracks with millisecond precision, and
• export a native CapCut “draft” project ready for final editing and multi‑platform publishing.

The tool is aimed at creators of micro‑dramas, film‑style commentary, highlight reels, and other short‑form video content.

---

🛠️ Core Components

Component	Role	Key Tech
Rust Native Multi‑Agent Engine	Drives the workflow, schedules tasks, and runs the AI agents.	Rust 1.85+, Cargo workspace
Agents	Specialized AI workers:

• DirectorAgent – overall planning and user interaction
• VisualCriticAgent – FFmpeg‑based frame analysis
• ScreenwriterAgent – generates 0‑3 s hook and first‑person monologue, includes a self‑reflection loop for completion rate
• VoiceArtistAgent – selects voice style, performs zero‑shot cloning (Edge‑TTS / GPT‑SoVITS)
• SoundEngineerAgent – 5‑track millisecond‑level alignment, BGM smart‑avoidance
• QualityReviewerAgent – checks timing, profanity, draft integrity | Rust crates under src‑tauri/crates/agent | | Tauri 2 Desktop Shell | Packages the app for macOS/Windows/Linux and provides native IPC between the front‑end and Rust back‑end. | Tauri 2, Rust | | React 19 Front‑End | Three‑pane “Integrated Cinema Studio” UI with four “workshops”: Asset Hub, Script Studio, Workspace, Export Hub. | React 19, TypeScript 5.8 | | LLM & TTS Integration | Connects to 11 LLM providers (OpenAI, Claude, Gemini, Qwen, DeepSeek, etc.) and multiple TTS / voice‑cloning services (Edge‑TTS, GPT‑SoVITS, local Ollama models). | HTTP clients, SDK wrappers | | Export Layer | Generates a .draft file compatible with CapCut/剪映 for seamless hand‑off to the final editor. | Rust export crate |

---

🚀 How It Works (Typical Workflow)

1. Import & Asset Hub – User drops a long video; the VisualCriticAgent extracts scenes, key frames, and emotional heatmaps.
2. Script Studio – The ScreenwriterAgent drafts a 3‑second hook and a first‑person narration, then the VoiceArtistAgent clones a chosen voice and produces the audio.
3. Workspace – The SoundEngineerAgent aligns the generated voice, background music, and original audio on a 5‑track timeline with < 18 ms drift. The QualityReviewerAgent validates the draft.
4. Export Hub – The final project is exported as a CapCut .draft and can be pushed to eight pre‑configured short‑video platforms (Douyin, Kuaishou, TikTok, YouTube, etc.) with built‑in duplicate‑content detection.

---

📦 Getting Started

# Clone the repo
git clone https://github.com/Agions/dramaxis.git
cd dramaxis

# Install front‑end deps (pnpm required)
pnpm install

# Run the desktop app in development mode
pnpm tauri dev

# Run the test suites
cargo test --manifest-path src-tauri/Cargo.toml --all
pnpm test

macOS tip – If the app is blocked as “damaged”, clear the quarantine flag:

sudo xattr -rd com.apple.quarantine /Applications/Dramaxis.app

---

📚 Documentation & License

• Full docs are in the docs/ folder and served via VitePress.
• The project is released under the MIT License.

---

🎯 Who Should Use It?

• Short‑form video creators who want AI‑assisted script writing, voice‑over generation, and precise editing.
• Social‑media managers looking to automate multi‑platform publishing with built‑in duplicate‑content checks.
• Developers interested in a Rust‑centric multi‑agent architecture for multimodal media pipelines.

---

🌐 Links

• Releases: https://github.com/Agions/dramaxis/releases
• Project website / docs: (served from docs/ via VitePress)
• License: MIT (see LICENSE)