Local Video Gen Studio

Generate stunning AI video clips from text—no subscriptions, no cloud fees, no waiting. Just your GPU and total creative control.

What is this?

Local Video Gen Studio is a cross-platform desktop application that brings professional video generation to your machine. Built around SANA-WM (a 2.6B open-source video model), it lets creators generate 720p video clips from text prompts in minutes. Process batches overnight, export to your editing timeline, and keep everything private—no cloud, no surveillance, no vendor lock-in.

Perfect for YouTubers building B-roll libraries, course creators illustrating concepts, and indie game developers prototyping visual effects without licensing costs.

Features

• Text-to-Video Generation – Convert detailed prompts into 720p video clips up to 60 seconds
• Batch Processing – Queue multiple videos and render overnight while you sleep
• Local-Only Inference – Your GPU, your data, your privacy. No cloud uploads
• GPU Auto-Detection – Automatically optimizes for NVIDIA/AMD/Metal
• Prompt Templates – Pre-built templates for common use cases (product demos, tutorials, game footage)
• One-Click Model Download – Automatic SANA-WM model initialization on first launch
• Export Ready – Direct MP4 output compatible with any video editor
• Cross-Platform – macOS, Windows, and Linux via Tauri

Quick Start

Requirements

• 12GB+ VRAM (NVIDIA RTX 3060 Ti / RTX 4070 or equivalent)
• Python 3.10+
• Node.js 16+
• 50GB free disk space (for model + cache)

Installation

1. Clone and install dependencies

   git clone https://github.com/yourusername/local-video-gen-studio.git
   cd local-video-gen-studio
   
   # Frontend
   npm install
   
   # Backend
   python -m venv venv
   source venv/bin/activate  # Windows: venv\Scripts\activate
   pip install -r backend/requirements.txt
   

2. Configure environment

   cp .env.example .env
   # Edit .env with your NVIDIA_API_KEY (optional) and output paths
   

3. Launch the app

   npm run tauri dev
   

The app will download SANA-WM (~3.2GB) on first run and auto-detect your GPU.

Usage

Generate a Single Video

1. Open Local Video Gen Studio
2. Enter a detailed text prompt (e.g., "A silver MacBook Pro sliding across a wooden desk, warm lighting, cinematic")
3. Select resolution (480p free tier, 720p Pro)
4. Click Generate
5. Monitor progress in the queue panel
6. Export to MP4 when complete

Batch Generate Variations

1. Create a prompt template with variables: "A {color} {object} {action}, {lighting}"
2. Upload a CSV with variations
3. Click Queue All – generates overnight
4. Retrieve finished videos in the exports folder

Example Prompts

"Overhead shot of coffee pouring into a white ceramic mug, steam rising, warm sunlight"

"Low-angle tracking shot through a forest canopy, autumn leaves, misty morning light"

"Screen recording aesthetic: cursor clicking through a minimalist dashboard, smooth transitions"

Tech Stack

Frontend:

• TypeScript + React (Vite)
• Tauri (lightweight Rust native shell)
• CSS3 for responsive UI

Backend:

• Python 3.10+
• SANA-WM (2.6B parameter open-source video model)
• PyTorch/CUDA for inference
• FastAPI for IPC
• SQLite for job queue persistence

Infrastructure:

• Stripe (optional monetization integration)
• Local GPU orchestration (no external services)

Development

# Run development server (hot reload)
npm run tauri dev

# Build production binaries
npm run tauri build

# Backend development (watch mode)
cd backend && python main.py --dev

# Type checking
npm run type-check

# Lint
npm run lint

Roadmap

• Frame interpolation (60fps output)
• LoRA fine-tuning for style consistency
• Realtime preview in editor
• Audio sync generation
• Commercial license watermark removal

Licensing

MIT License – See LICENSE file for details.

Questions? Open an issue or check out OVERVIEW.md for deeper architecture docs.