Spaces:
Running
Running
π» GhostPackDemo: Next-Gen AI Video Generator
Create real 30 FPS AI videos from a single image on your own GPU
π What is GhostPackDemo?
GhostPackDemo is an open-source, GPU-optimized AI video generator. It fuses HunyuanVideo and FramePack into a single streamlined pipeline with a beautiful Gradio UI.
- Ultra Fast: Up to 40% faster thanks to Teacache, Sage-Attention, and dynamic batching.
- Resource-Smart: 50% lower VRAM (context packing, tcmalloc, memory cache).
- Laptop-Friendly: Works with 6GB VRAM, 8GB RAM (e.g., GTX 1650, RTX 3050/3060).
- No Cloud Needed: Run on your local desktop, server, or laptop.
- Full Control: Live preview, prompt/batch/frame/CRF settings, instant exports.
β¨ Features
- Veo 3-Level AI: Next-frame video prediction for photorealistic motion.
- Phantom Speed: Teacache & Sage-Attention = 12β15s/frame (RTX 3060).
- Batch & Prompt Tuning: Full workflow control, see frames build in real time.
- Open Source: Apache 2.0, remix, extend, contribute.
- Export Quality: MP4 with customizable CRF (0β100).
π§ Optimizations & Requirements
- Speed
- Teacache: ~40% faster (
T_total β 0.6 Γ T_base
) - Sage-Attention: +10%
- CUDA tweaks: β15% latency
- Dynamic scheduling: β20% time
- Teacache: ~40% faster (
- Memory
- Context packing: β50% VRAM
- tcmalloc: β20% overhead
- Memory cache: β25% VRAM swap
- Compute
- Dynamic batching: +50% throughput
- Efficiency
- Power saving: β30% idle power
- Thread tuning: +15% CPU
- VRAM Needs
- GTX 1650 (6GB): 18β25s/frame
- RTX 3050 (8GB): 15β20s/frame
- RTX 3060 (12GB): 10β15s/frame
- RTX 4090 (24GB): 1.5β2.5s/frame
Minimum: NVIDIA GPU (6GB+ VRAM), 8GB RAM, 30GB free disk, Python 3.10+, CUDA 12.6+
Optimal: RTX 3060 or higher
πΊ Demo & Screenshots




See more: Demo Videos
π οΈ Installation
Ubuntu
git clone https://github.com/ghostai1/GhostPack
cd GhostPack
chmod +x install_ubuntu.sh
./install_ubuntu.sh
Windows
git clone https://github.com/ghostai1/GhostPack
cd GhostPack
install.bat
macOS
git clone https://github.com/ghostai1/GhostPack
cd GhostPack
chmod +x install_macos.sh
./install_macos.sh
π¦ Quick Start
source ~/ghostpack_venv/bin/activate
cd ~/ghostpack_venv
python ghostgradio.py --port 5666 --server 0.0.0.0
- Upload an image and enter your prompt
- Set FPS, batch size, video length, CRF, enable Teacache
- Watch your video build in real-time preview
- Export to MP4, any quality
- Monitor GPU usage:
nvidia-smi
π Roadmap
- AI Voice Generation: Narrate videos with custom voices
- AI Start Images: Generate ultra-high-res starting frames
- Native VR: Experimental 3D/VR video support
π€ Community & Support
- Discord β Support & collabs
- GitHub Issues β Report bugs & request features
- Hugging Face Space β Live web demo
π§βπ» Contributing
- Fork, branch, PR. All improvements welcome!
- See CONTRIBUTING.md
π License
Apache 2.0
GhostPack by ghostai1 Β· Hugging Face
Created June 11, 2025