REPLICATE_SETUP.md

🚀 Using Replicate API for Real Hailuo (MiniMax) Video Generation

If Hugging Face Spaces are not working, you can use Replicate API which has the actual Hailuo video-01 model by MiniMax!

🌟 Why Replicate?

• ✅ Real Hailuo Model: The actual MiniMax video-01 model
• ✅ Reliable: Always available, no sleeping spaces
• ✅ Fast: Optimized infrastructure
• ✅ Free Tier: $5 free credits to start

📝 Setup Instructions

1. Get Replicate API Token

1. Go to https://replicate.com
2. Sign up for a free account
3. Go to https://replicate.com/account/api-tokens
4. Copy your API token

2. Install Replicate Python Client

pip install replicate

3. Create Replicate Backend

Create a new file backend_replicate.py:

from flask import Flask, request, jsonify
from flask_cors import CORS
import replicate
import os
from dotenv import load_dotenv
from datetime import datetime
import logging

load_dotenv()

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

app = Flask(__name__)
CORS(app)

# Set your Replicate API token
REPLICATE_API_TOKEN = os.getenv('REPLICATE_API_TOKEN')
if REPLICATE_API_TOKEN:
    os.environ['REPLICATE_API_TOKEN'] = REPLICATE_API_TOKEN

@app.route('/health', methods=['GET'])
def health():
    return jsonify({'status': 'healthy', 'service': 'Replicate API'})

@app.route('/generate-video', methods=['POST'])
def generate_video():
    try:
        data = request.json
        prompt = data.get('prompt', '')
        
        if not prompt:
            return jsonify({'error': 'Prompt is required'}), 400
        
        logger.info(f"Generating video with Hailuo: {prompt[:100]}")
        
        # Use the real Hailuo (MiniMax) model on Replicate
        output = replicate.run(
            "minimax/video-01",
            input={
                "prompt": prompt,
                "prompt_optimizer": True
            }
        )
        
        # Output is a video URL
        video_url = output if isinstance(output, str) else output[0]
        
        logger.info(f"Video generated: {video_url}")
        
        return jsonify({
            'video_url': video_url,
            'prompt': prompt,
            'model': 'hailuo-minimax',
            'model_name': 'Hailuo Video-01 (MiniMax)',
            'timestamp': datetime.now().isoformat()
        })
        
    except Exception as e:
        logger.error(f"Error: {str(e)}")
        return jsonify({'error': str(e)}), 500

if __name__ == '__main__':
    if not REPLICATE_API_TOKEN:
        print("⚠️  Warning: REPLICATE_API_TOKEN not set in .env file")
        print("Get your token from: https://replicate.com/account/api-tokens")
    
    app.run(host='0.0.0.0', port=5000, debug=False)

4. Update .env File

Add your Replicate API token to .env:

REPLICATE_API_TOKEN=r8_your_token_here

5. Run the Replicate Backend

python3 backend_replicate.py

6. Use with Your Frontend

The frontend (index_enhanced.html) will work with this backend without any changes!

🎬 Available Models on Replicate

Text-to-Video Models:

1. Hailuo Video-01 (MiniMax) - The real Hailuo model!

   replicate.run("minimax/video-01", input={"prompt": "..."})
   

2. CogVideoX

   replicate.run("lucataco/cogvideox-5b", input={"prompt": "..."})
   

3. Kling v1.6

   replicate.run("fofr/kling-v1.6-standard", input={"prompt": "..."})
   

Image-to-Video Models:

1. Stable Video Diffusion

   replicate.run("stability-ai/stable-video-diffusion", input={"image": "..."})
   

2. Hailuo Image-to-Video

   replicate.run("minimax/video-01", input={"image": "...", "prompt": "..."})
   

💰 Pricing

• Free Tier: $5 free credits (enough for ~50 videos)
• Pay-as-you-go: ~$0.10 per video generation
• Much cheaper than running your own GPU!

🔧 Advanced: Multi-Model Support

Create backend_replicate_multi.py with multiple models:

MODELS = {
    "hailuo": "minimax/video-01",
    "cogvideox": "lucataco/cogvideox-5b",
    "kling": "fofr/kling-v1.6-standard",
    "svd": "stability-ai/stable-video-diffusion"
}

@app.route('/generate-video', methods=['POST'])
def generate_video():
    data = request.json
    model_id = data.get('model', 'hailuo')
    prompt = data.get('prompt', '')
    
    model_name = MODELS.get(model_id, MODELS['hailuo'])
    
    output = replicate.run(model_name, input={"prompt": prompt})
    # ... rest of the code

✅ Advantages of Replicate

1. Real Hailuo Model: The actual MiniMax video-01
2. No Space Sleeping: Always available
3. Better Performance: Optimized infrastructure
4. Predictable Costs: Pay per generation
5. Multiple Models: Access to many models
6. Better Support: Professional API support

📚 Resources

• Replicate Docs: https://replicate.com/docs
• Hailuo Model: https://replicate.com/minimax/video-01
• Python Client: https://github.com/replicate/replicate-python
• Pricing: https://replicate.com/pricing

🎯 Recommendation

For production use or if Hugging Face Spaces keep failing:

• ✅ Use Replicate API
• ✅ Get the real Hailuo model
• ✅ Reliable and fast
• ✅ Only pay for what you use

For testing/development:

• Use Demo Mode in the current app
• Or use Hugging Face Spaces (free but may be slow/unavailable)

---

Ready to use the real Hailuo model? Set up Replicate now! 🚀