High-fidelity AI Text-to-Video and Image-to-Video generation studio featuring a Gradio interface and zero-cost cloud GPU worker bridge.
Features • Architecture • Cloud Acceleration • Installation • License
VideoStudio is a state-of-the-art AI video synthesis suite that enables creators, developers, and researchers to generate cinematic 1080p and 4K video clips directly from text prompts or still images. Engineered in Python with a rich Gradio UI, VideoStudio bridges local workstation controls with free cloud GPU backends (e.g. Google Colab T4/A100) to render state-of-the-art diffusion models (Wan 2.1, LTX-Video) without requiring a multi-thousand-dollar local GPU.
- 🎬 Multi-Modal Video Synthesis:
- Text-to-Video (T2V): Generates fluid, temporally coherent video sequences from descriptive natural language prompts.
- Image-to-Video (I2V): Animates still photographs, digital art, or AI portraits with cinematic motion trajectories.
- ⚡ Free Cloud GPU Acceleration (
colab_worker.ipynb): Offload massive VRAM compute requirements to free Google Colab cloud instances via automated tunneling (Ngrok / Gradio link), keeping your local machine completely unburdened. - 🎨 Modern Gradio Interactive Studio:
- Resolution controls (480p, 720p, 1080p)
- Frame rate configuration (24fps, 30fps) and duration sliders
- Guidance scale, motion bucket ID, and seed randomization
- Built-in video preview player with instant download
- 🖥️ Turnkey Windows Launch: Pre-configured
Start Video Studio.batand silentrun_videostudio.vbsfor one-click desktop initiation.
flowchart LR
A[Creator: Text Prompt / Image] --> B[Local Gradio Studio: app.py]
B -->|Encrypted Remote Tunnel| C[Cloud GPU Worker: colab_worker.ipynb]
subgraph Cloud Diffusion Engine
C --> D[Wan 2.1 / LTX-Video Pipeline]
D --> E[Temporal Attention & VAE Decoder]
E --> F[FFmpeg MP4 Encode]
end
F -->|Return Video Stream| B
B --> G[(Local Video Export & Preview Player)]
VideoStudio/
├── app.py # Primary Gradio web studio interface
├── video_studio.py # Video generation pipeline orchestrator
├── colab_worker.ipynb # Cloud GPU worker notebook for free acceleration
├── Wan2.1-main/ # Integrated Wan2.1 diffusion model codebase
├── Start Video Studio.bat # Windows one-click desktop launcher
├── run_videostudio.vbs # Silent background VBS launcher
├── videostudio.ico # High-resolution application icon
├── requirements.txt # Python dependencies
├── .gitignore # Video cache and weights exclusions
└── LICENSE # Open-source MIT License
- Python 3.10 or higher
- (Optional for local inference) NVIDIA GPU with 12GB+ VRAM or free Google Colab account
git clone https://github.com/Kamran5H/VideoStudio.git
cd VideoStudio
# Setup virtual environment
python -m venv .venv
.venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt# Start local studio
python app.py
# Or double-click "Start Video Studio.bat" on WindowsRun the offline verification suite with:
python -m unittest discover -s tests -vtest_gemini_features.py and test_verify.py are opt-in live smoke checks; they may use network services, API quota, and local media processing, so they are not run during automated test discovery.
Submitted video jobs run in the local studio worker, not in the browser page. Refreshing the page reconnects to the saved job queue and restores its status and completed video; the current phase, queue position, or quota retry window is shown in the shared status panel above the tabs. Active work uses an indeterminate progress bar rather than an invented percentage or ETA. When no job is running, the latest completed result is reopened in the tab that created it.
Creative form values and the selected tab are auto-saved in the current browser tab's session storage, so they survive a page refresh but are cleared when that browser tab session ends. Unsaved API-key edits are never stored in browser state. Use Clear saved draft to reset creative fields without cancelling jobs or removing videos.
This project is open-source and released under the MIT License.
Copyright (c) 2024-2026 Kamran Ashraf.