QUICK_START.md

# Quick Start Guide

## 🚀 Deploy in 5 Minutes

### Step 1: Complete the Inference (REQUIRED)
⚠️ **The code has placeholders for actual video generation**

See [TODO.md](./TODO.md) for details on integrating the inference logic.

### Step 2: Create HuggingFace Space

1. Go to https://huggingface.co/new-space
2. Fill in:
   - **Name**: `infinitetalk` (or your choice)
   - **License**: `apache-2.0`
   - **SDK**: `Gradio`
   - **Hardware**: `ZeroGPU` ✨ (FREE tier available!)
3. Click **Create Space**

### Step 3: Upload Files

**Via Web UI** (easiest):
1. Click "Files" tab in your Space
2. Drag and drop all files from this directory:
   ```
   README.md
   app.py
   requirements.txt
   packages.txt
   .gitignore
   LICENSE.txt
   src/ (folder)
   wan/ (folder)
   utils/ (folder)
   assets/ (folder)
   examples/ (folder)
   ```
3. Click "Commit changes"

**Via Git**:
```bash
git clone https://huggingface.co/spaces/YOUR_USERNAME/YOUR_SPACE_NAME
cd YOUR_SPACE_NAME
cp -r /path/to/infinitetalk-hf-space/* .
git add .
git commit -m "Initial deployment"
git push
```

### Step 4: Wait for Build

- Build time: **5-10 minutes**
- Check "Logs" tab for progress
- Flash-attn compilation takes longest

### Step 5: Test

1. Space shows "Running" ✅
2. First generation downloads models (2-3 min)
3. Try image-to-video example
4. Try video dubbing example

## ⚡ Quick Commands

```bash
# View directory structure
ls -la

# Check file sizes
du -sh *

# Count lines of code
find . -name "*.py" | xargs wc -l

# Test Python syntax
python -m py_compile app.py

# View logs (after deployment)
# Go to your Space → Logs tab
```

## 🎯 Common Issues & Fixes

### Build Fails
- **Check Logs tab** for specific error
- **Flash-attn timeout?** Normal, wait 10-15 min
- **Still failing?** Try Dockerfile approach (see DEPLOYMENT.md)

### Models Don't Download
- Check https://status.huggingface.co
- Verify model repo IDs in `utils/model_loader.py`
- Add HF_TOKEN in Space settings if needed

### Out of Memory
- Use 480p instead of 720p
- Reduce steps to 30
- Process shorter videos (<10s)

### Space Stuck
- Refresh page
- Check if in queue (ZeroGPU)
- Wait for quota to refill

## 📊 Files Overview

| File/Folder | Purpose | Lines | Critical? |
|-------------|---------|-------|-----------|
| `README.md` | Space metadata | ~50 | ✅ Yes |
| `app.py` | Main application | ~350 | ✅ Yes |
| `requirements.txt` | Python packages | ~30 | ✅ Yes |
| `packages.txt` | System packages | ~4 | ✅ Yes |
| `utils/model_loader.py` | Model management | ~200 | ✅ Yes |
| `utils/gpu_manager.py` | Memory management | ~150 | ✅ Yes |
| `src/` | Audio analysis | - | ✅ Yes |
| `wan/` | Model code | - | ✅ Yes |
| `assets/` | UI assets | - | Optional |
| `examples/` | Sample data | - | Optional |

## 🔧 Pre-Deployment Checklist

- [x] All files present
- [x] README.md has YAML metadata
- [x] requirements.txt is properly ordered
- [x] ZeroGPU hardware configured
- [ ] **Inference logic integrated** ⬅️ CRITICAL
- [ ] Tested locally (if possible)
- [ ] Examples prepared

## 💰 Cost Breakdown

### Free Tier
- **Cost**: $0
- **GPU**: H200 (70GB VRAM)
- **Quota**: 300s per session, 600s max
- **Usage**: ~3-5 generations per quota
- **Best for**: Testing, demos, light use

### PRO Tier
- **Cost**: $9/month
- **GPU**: Same H200
- **Quota**: 8× more (1500s)
- **Spaces**: Up to 10
- **Best for**: Regular use, public demos

## 📈 Performance Expectations

| Task | Resolution | Time | VRAM |
|------|-----------|------|------|
| Model download | - | 2-3 min | - |
| 10s video | 480p | ~40s | ~38GB |
| 10s video | 720p | ~70s | ~55GB |
| 30s video | 480p | ~90s | ~45GB |

## 🎓 Learning Resources

- [HuggingFace Spaces Tutorial](https://huggingface.co/docs/hub/spaces-overview)
- [Gradio Documentation](https://gradio.app/docs)
- [ZeroGPU Guide](https://huggingface.co/docs/hub/spaces-zerogpu)
- [InfiniteTalk Paper](https://arxiv.org/abs/2508.14033)

## ✅ Success Checklist

After deployment:

1. [ ] Space builds successfully
2. [ ] No errors in Logs
3. [ ] UI loads properly
4. [ ] Models download on first run
5. [ ] Image-to-video works
6. [ ] Video dubbing works
7. [ ] No OOM errors
8. [ ] Memory cleanup works
9. [ ] Can run multiple generations
10. [ ] Results look good!

## 🆘 Need Help?

1. **Check** [TODO.md](./TODO.md) for implementation details
2. **Read** [DEPLOYMENT.md](./DEPLOYMENT.md) for troubleshooting
3. **Review** [PROJECT_SUMMARY.md](./PROJECT_SUMMARY.md) for overview
4. **Ask** on HuggingFace Forums: https://discuss.huggingface.co
5. **File issue** on InfiniteTalk GitHub: https://github.com/MeiGen-AI/InfiniteTalk

---

**Ready?** Complete the inference integration, then deploy! 🚀