Hugging Face's logo

sravya
/
videoAI

Model card Files
xet
Community
videoAI / ONLINE_VS_LOCAL.md
sravya's picture
sravya
Upload 33 files
54ed165 verified
preview code
|
raw
history blame contribute delete
2.8 kB

🌐 Online Models vs Local Models

✅ Now Using: Online Models (RECOMMENDED)

You're now using backend_enhanced.py which connects to free Hugging Face Spaces!

🚀 Available Online Models:

  1. CogVideoX-5B (Best Quality) ⭐

    • 6 seconds, 720p
    • Generation time: 30-60 seconds
    • Free on Hugging Face Zero GPU

  2. CogVideoX-2B (Faster)

    • 6 seconds, 720p
    • Generation time: 20-40 seconds
    • Good quality, faster

  3. HunyuanVideo (Tencent - SOTA)

    • State-of-the-art quality
    • Longer videos possible
    • Generation time: 60-90 seconds

  4. Demo Mode (Instant Testing)

    • Returns sample video immediately
    • Perfect for testing the UI

✨ Advantages of Online Models:

✅ No downloads - No 5GB model to download ✅ Fast - 30-60 seconds per video (vs 5-10 minutes on CPU) ✅ No GPU needed - Runs on Hugging Face's servers ✅ Better quality - Access to larger, better models ✅ Multiple models - Switch between different AI models ✅ Camera controls - Hailuo-inspired camera movements ✅ Visual effects - Cinematic, dramatic, slow-motion, etc.

⚠️ Limitations:

  • Requires internet connection
  • May have queue times if many people are using it
  • Some spaces may be sleeping (takes 30s to wake up)

🖥️ Local Model (backend_local.py)

When to Use Local:

  • ✅ You have a powerful NVIDIA GPU (RTX 3060+)
  • ✅ You want 100% privacy (no data sent to cloud)
  • ✅ You don't have reliable internet
  • ✅ You want to run custom models

Disadvantages:

  • ❌ 5GB download required
  • ❌ Very slow on CPU (5-10 minutes per video)
  • ❌ Requires 16GB+ RAM
  • ❌ Only one model available
  • ❌ No advanced features

📊 Comparison Table

Feature Online (Enhanced) Local
Setup Time Instant 10-30 minutes
Download Size 0 GB 5 GB
Generation Speed 30-60 sec 5-10 min (CPU)
Quality Excellent Good
Models Available 4+ models 1 model
Camera Controls ✅ Yes ❌ No
Visual Effects ✅ Yes ❌ No
GPU Required ❌ No ⚠️ Recommended
Internet Required ✅ Yes ❌ No
Privacy Cloud-based 100% local
Cost Free Free

🎯 Recommendation

Use Online Models (backend_enhanced.py) unless you specifically need:

  • Complete privacy
  • Offline generation
  • Custom model training

🔄 How to Switch

To Online (Current): 

/Users/sravyalu/VideoAI/.venv/bin/python backend_enhanced.py
# Open: index_enhanced.html

To Local: 

/Users/sravyalu/VideoAI/.venv/bin/python backend_local.py
# Open: index_local.html

You're all set with online models! Much faster and easier! 🎉