Spaces:

ghostai1
/

GhostPack

Running

App Files Files Community

ghostai1 commited on Jun 12

Commit

8f926f1

verified ·

1 Parent(s): e4fd977

Update GhostPackDemo/readme.md

Browse files

Files changed (1) hide show

GhostPackDemo/readme.md +78 -116

GhostPackDemo/readme.md CHANGED Viewed

@@ -1,71 +1,83 @@
-# 👻🎬 GhostPackDemo
-**The fastest open-source AI video generator for local GPUs (Apache-2.0)**
-Create stunning **30 FPS** AI videos from a single image with advanced local GPU acceleration.
 [![Python](https://img.shields.io/badge/python-3.8%2B-blue?logo=python)](https://python.org)
 [![License](https://img.shields.io/badge/license-Apache%202.0-green)](LICENSE)
-[![HF Space](https://img.shields.io/badge/🤗%20Spaces-Live%20Demo-orange)](https://huggingface.co/spaces/ghostai1/GhostPackDemo)
 ---
-## 🚀 Overview
-GhostPackDemo merges **HunyuanVideo** and **FramePack** into a seamless GPU video pipeline (`ghostpack_main_fixed.py`) with a powerful, easy-to-use **Gradio** interface.
-- **Run on nearly any modern GPU with 6GB+ VRAM.**
-- **Photorealistic, next-gen AI video from a single image.**
-- **Perfect for laptops (8GB RAM) and desktops.**
 ---
-## ✨ Key Features
-- **Veo 3-Level AI**: Next-frame prediction for hyper-realistic videos.
-- **Phantom Speed**: Teacache & Sage-Attention make it 40–50% faster than base.
-- **Flexible**: Control frame length, batch size, and creative prompts.
-- **Live Preview**: See your video build in real time.
-- **Community Open Source**: Apache 2.0 license, full repo, no paywalls.
 ---
-## 🚀 Optimizations
-- **Speed**
-    - Teacache: +40% (`T_total ≈ 0.6 × T_base`)
-    - Sage-Attention: +10%
-    - CUDA Tweaks: –15% latency
-    - Dynamic Scheduling: –20% generation time
-- **Memory**
-    - Context Packing: –50% VRAM use
-    - tcmalloc: –20% overhead
-    - Memory Cache: –25% VRAM swaps
-- **Compute**
-    - Dynamic Batching: +50% throughput
-- **Efficiency**
-    - Power Saving: –30% idle power
-    - Thread Tuning: +15% CPU boost
-- **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
 ---
-## 📺 Demo
-- **Watch a neon-lit spectral battle come to life.**
-- [More Demo Videos](https://github.com/ghostai1/GhostPack/blob/main/demo_videos)
 ---
 ## 🛠️ Installation
-**Requires:**
-- NVIDIA GPU (6GB+ VRAM), 8GB+ RAM, >30GB free disk
-- Python 3.10, CUDA 12.6
 ### Ubuntu
 ```bash
 git clone https://github.com/ghostai1/GhostPack
 cd GhostPack
@@ -74,7 +86,6 @@ chmod +x install_ubuntu.sh
 ```
 ### Windows
 ```bash
 git clone https://github.com/ghostai1/GhostPack
 cd GhostPack
@@ -82,7 +93,6 @@ install.bat
 ```
 ### macOS
 ```bash
 git clone https://github.com/ghostai1/GhostPack
 cd GhostPack
@@ -92,95 +102,47 @@ chmod +x install_macos.sh
 ---
-## ⚡ How To Use
-1. **Start the GUI:**
-    ```bash
-    source ~/ghostpack_venv/bin/activate
-    cd ~/ghostpack_venv
-    python ghostgradio.py --port 5666 --server 0.0.0.0
-    ```
-2. **Generate a video:**
-    - Upload a source image.
-    - Enter a prompt, e.g., “A spectral figure glides through a neon-lit city.”
-    - Set FPS, length, and enable Teacache.
-    - Click **Generate**.
-3. **Monitor GPU:**
-    Use `nvidia-smi` to check for 80–100% GPU usage.
-4. **Export:**
-    Save as MP4 with your chosen quality.
----
-## 🔮 Roadmap
-- **AI Voice**: Narrated videos
-- **AI Images**: Create start frames
-- **VR**: Experimental 3D/VR video
----
-## 🌎 Community & Support
-- [Discord](#) — Chat, share, get help
-- [Submit Issues](#) — Bug reports, requests
 ---
-## 📖 Technical Deep Dive
-- **Core:** HunyuanVideo + FramePackI2V_HY + LlamaModel + CLIPTextModel
-- **Diffusion:** K-Diffusion, Unipc pipeline
-- **Docs:** [Wiki](#)
 ---
-## ❓ FAQ
-- **Q:** Minimum hardware?
-  **A:** 6GB+ VRAM, 8GB+ RAM, CUDA GPU
-- **Q:** How do I install?
-  **A:** Clone repo, run install script for your OS.
-- **Q:** Can I contribute?
-  **A:** Yes! Fork, branch, PR. See guidelines in repo.
 ---
 ## 🧑‍💻 Contributing
-- Fork & branch the repo
-- Submit pull requests
-- Follow wiki/test with `ghostgradio.py`
-- [Contribution Guidelines](#)
----
-## 🖼️ Screenshots
-**Main UI:**
-![Main Interface](assets/screenshots/main_interface.png)
-**Advanced Settings:**
-![Advanced Settings](assets/screenshots/advanced_settings.png)
-**Logs:**
-![Logs Display](assets/screenshots/logs_display.png)
----
-## 🗂️ Latest Files
-- **ghostgradio.py**
-- [Logs](#)
-- [Refresh Repo](#) | [Refresh Logs](#)
 ---
 ## 📜 License
-Apache 2.0. See [LICENSE](LICENSE).
----
-**👻 GhostPack by [ghostai1](https://github.com/ghostai1/GhostPack) | [Hugging Face](https://huggingface.co/ghostai1)**
 *Created June 11, 2025*

+# 👻 GhostPackDemo: Next-Gen AI Video Engine
+**Create real 30FPS AI videos from a single image on any modern GPU!**
 [![Python](https://img.shields.io/badge/python-3.8%2B-blue?logo=python)](https://python.org)
 [![License](https://img.shields.io/badge/license-Apache%202.0-green)](LICENSE)
+[![Live Demo](https://img.shields.io/badge/🤗%20Spaces-Live%20Demo-orange)](https://huggingface.co/spaces/ghostai1/GhostPackDemo)
 ---
+## 🚀 What is GhostPackDemo?
+GhostPackDemo is a **laptop-friendly, open-source AI video generator** that merges HunyuanVideo and FramePack into one supercharged pipeline with an intuitive Gradio UI.
+No cloud, no rental GPUs: **make stunning, smooth videos on your own hardware** (6GB+ VRAM/8GB+ RAM).
+- **Blazing Fast**: Up to 40% faster with Teacache & Sage-Attention
+- **Efficient**: 50% less VRAM required (context packing, batching, tcmalloc, cache tricks)
+- **Customizable**: Control every parameter, live preview, tune batch size & prompts
+- **Real Results**: 10–15s/frame (RTX 3060), full 30FPS from one photo
+- **Fully Open Source**: Apache 2.0
 ---
+## ✨ Features
+- **Veo 3-Level AI**: Advanced frame prediction for lifelike motion
+- **Phantom Speed**: 12–15s/frame (RTX 3060), up to 50% throughput boost with dynamic batching
+- **Laptop & Desktop Ready**: Works on 6GB/8GB+ VRAM (GTX 1650 and up)
+- **Custom Workflow**: Batch, frame count, prompt, CRF—dial it all in
+- **Instant Preview**: See frames as they're built (Gradio)
+- **Community Built**: Free, extensible, remixable
 ---
+## 🧙‍♂️ Math Sorcery (Optimizations)
+- **Speed**:
+  - Teacache: ~40% faster (`T_total ≈ 0.6 × T_base`)
+  - Sage-Attention: 5–10% faster
+  - CUDA tweaks: –15% latency
+  - Dynamic scheduling: –20% gen time
+- **Memory**:
+  - Context packing: –50% VRAM
+  - tcmalloc: –20% overhead
+  - VRAM cache: –25% swap ops
+- **Compute**:
+  - Dynamic batching: +50% throughput
+- **Efficiency**:
+  - Power savings: –30% idle draw
+  - 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
 ---
+## 📺 Demos & Screenshots
+<div align="center">
+<img src="https://cdn-uploads.huggingface.co/production/uploads/6421b1c68adc8881b974a89d/k8pgUlg4OvdUZpbMNTcp5.gif" alt="GhostPack Demo GIF" style="max-width: 480px; border-radius: 14px; box-shadow: 0 0 30px #ff00ff; margin-bottom: 20px;">
+<br>
+<img src="https://cdn-uploads.huggingface.co/production/uploads/6421b1c68adc8881b974a89d/7ABE2lOA4LOUtPfh1mhxP.png" alt="Main Interface" width="310" style="margin:12px; border-radius:12px; box-shadow:0 0 16px #00ffcc;">
+<img src="https://cdn-uploads.huggingface.co/production/uploads/6421b1c68adc8881b974a89d/9qNVRX2eM2iCY8xQKcOwW.png" alt="Advanced Settings" width="310" style="margin:12px; border-radius:12px; box-shadow:0 0 16px #00ffcc;">
+<img src="https://cdn-uploads.huggingface.co/production/uploads/6421b1c68adc8881b974a89d/--fIS9ITg4-VqN22ySoa2.png" alt="Logs Display" width="310" style="margin:12px; border-radius:12px; box-shadow:0 0 16px #00ffcc;">
+</div>
+> **Main Interface · Advanced Settings · Logs**
+See more: [Demo Videos](https://github.com/ghostai1/GhostPack/blob/main/demo_videos)
 ---
 ## 🛠️ Installation
+**Requires:** NVIDIA GPU (6GB+ VRAM), 8GB+ RAM, 30GB+ disk, Python 3.10, CUDA 12.6+
 ### Ubuntu
 ```bash
 git clone https://github.com/ghostai1/GhostPack
 cd GhostPack
 ```
 ### Windows
 ```bash
 git clone https://github.com/ghostai1/GhostPack
 cd GhostPack
 ```
 ### macOS
 ```bash
 git clone https://github.com/ghostai1/GhostPack
 cd GhostPack
 ---
+## 🚦 Quick Start
+```bash
+source ~/ghostpack_venv/bin/activate
+cd ~/ghostpack_venv
+python ghostgradio.py --port 5666 --server 0.0.0.0
+```
+- **Upload an image**, set your prompt, choose batch size/FPS/CRF.
+- **Enable Teacache** for best speed.
+- **Watch the live preview.**
+- **Export your video** (MP4, any quality).
+- **Monitor GPU:** run `nvidia-smi` for full load.
 ---
+## 🌟 Roadmap
+- **AI Voice Generation**: Narrate your videos with custom AI voices
+- **AI Image Creation**: Start with pro-quality AI-generated frames
+- **VR Video**: Native support for 3D/VR video rendering
 ---
+## 🤝 Community
+- [Discord](https://discord.gg/ghostpack) — Support, collabs, & feedback
+- [GitHub Issues](https://github.com/ghostai1/GhostPack/issues) — Report bugs, request features
+- [Hugging Face Space](https://huggingface.co/spaces/ghostai1/GhostPackDemo) — Try the demo live
 ---
 ## 🧑‍💻 Contributing
+- Fork & branch, submit PRs, follow [CONTRIBUTING.md](https://github.com/ghostai1/GhostPack/blob/main/CONTRIBUTING.md)
+- All improvements welcome—GUI, docs, optimization, features
 ---
 ## 📜 License
+Apache 2.0.
+**GhostPack by [ghostai1](https://github.com/ghostai1/GhostPack) · [Hugging Face](https://huggingface.co/ghostai1)**
 *Created June 11, 2025*