👻🚧 GhostPackDemo (BETA) – Next-Gen AI Video Generator

Beta Demo — Try experimental AI video creation, directly on your GPU!

GhostPack Animated Banner

🚀 What is GhostPackDemo? (Beta)

GhostPackDemo is an open-source AI video generator for your local GPU. This BETA combines HunyuanVideo and FramePack into a single Python pipeline + live Gradio UI.

🦾 Fast: 40% faster with Teacache, Sage-Attention, CUDA tweaks.
🧠 Efficient: 50% less VRAM (context packing, tcmalloc, memory cache).
💻 Laptop-Ready: 6GB VRAM & 8GB RAM minimum (GTX 1650/RTX 3050+).
🔒 Runs Local: All processing on your own hardware — no cloud!
🎛️ Live Preview: See every frame build in real time, full workflow control.

Beta Notice: Expect rough edges, limited samplers, and edge-case bugs! This demo build prioritizes clarity and reproducibility for AI devs.

✨ Features (Beta)

Veo 3-Level AI: Next-frame prediction for ultra-realistic video motion.
Phantom Speed: Teacache + Sage-Attention, 12–15s/frame (RTX 3060).
Config Control: Batch size, frame window, prompt, CRF, see everything live.
Open Source: Apache 2.0 — fork, remix, contribute!
Export: High quality MP4 (CRF 0–100).

🧙 Math Sorcery & Hardware

Speed: Teacache ~40% (T_total ≈ 0.6 × T_base), Sage-Attention +10%, CUDA tweaks –15% latency
Memory: Context packing –50% VRAM, tcmalloc –20% overhead, memory cache –25%
Compute: Dynamic batching +50% throughput
Efficiency: Power save –30% idle, Thread tuning +15% CPU
VRAM Needs:
- GTX 1650 (6GB): 18–25s/frame
- RTX 3050 (8GB): 15–20s/frame
- RTX 3060 (12GB): 10–15s/frame
- RTX 4090 (24GB): 1.5–2.5s/frame

Minimum: NVIDIA GPU (6GB+), 8GB RAM, 30GB+ disk, Python 3.10+, CUDA 12.6+
Recommended: RTX 3060 or better

📑 .env Setup (IMPORTANT)

You must add your Hugging Face token to .env to download models and samplers!

Create a file called .env in the project root with:

HF_TOKEN=hf_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Get your token from: https://huggingface.co/settings/tokens
If missing, model download and pipeline will fail.

📺 Demo & Screenshots

_{Main Interface • Advanced Settings • Logs}

🎬 More Demo Clips

⚙️ Installation (Beta)

💾 30GB+ free space needed. Installs models on first run.

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

⚡ Quick Start

source ~/ghostpack_venv/bin/activate
cd ~/ghostpack_venv
python ghostgradio.py --port 5666 --server 0.0.0.0

🖼️ Upload an image
💬 Enter a prompt (e.g. “A graceful dance movement”)
🔄 Enable Teacache, set video seconds, steps, CRF
👀 See live frame preview + logs
💾 Export MP4 instantly
🖥️ Monitor GPU: nvidia-smi

🗺️ Roadmap

🗣️ AI Voice: Add local voiceover/narration
🖼️ AI Start Images: Next-level starting frames
🕶️ VR Support: Immersive AI video output
🔒 More Samplers/Models coming soon

💬 Community & Help

Discord – Feedback & support
GitHub Issues – Bugs & requests
Hugging Face Space – Try the live demo

👨‍💻 Contributing

Fork, branch, and PR! Beta testers wanted!
See CONTRIBUTING.md

🪪 License

Apache 2.0
GhostPack by ghostai1 · Hugging Face
BETA build · Created June 11, 2025