Not-Real-Konsistent: Consistent Character Video Generator

Generate realistic videos with consistent character preservation. This Replicate model specializes in maintaining character and object consistency from reference images while creating authentic-looking animated videos. Perfect for creating videos where character identity needs to remain consistent throughout.

🎯 What it does

This model excels at creating realistic videos while maintaining character consistency:

1. Consistent Character Generation: When provided with a reference image, the model preserves character identity and features throughout the video
2. Flexible Scene Creation: Generates new scenes and actions while keeping characters/objects consistent
3. Realistic Animation: Transforms images into natural-looking videos with authentic motion

🚀 Try it Online

Live Model: https://replicate.com/rinesh/not-real-konsistent

✨ Features

• Character Consistency:
  • Maintains character/object identity from reference images
  • Preserves key features and attributes throughout the video
  • Perfect for creating consistent character animations
• Customizable Output:
  • Duration: 5 or 10 second videos
  • Resolution: 480p or 1080p
  • Aspect Ratios: Multiple formats (16:9, 9:16, 1:1, 4:3, etc.)
• Dual Output: Get both the generated image and video
• Automatic Authentication: No API tokens needed when using on Replicate

📝 Input Parameters

Parameter	Type	Description	Default
input_image	File	Optional reference image for character/object consistency	None
prompt	String	Describe the scene/action (character consistency maintained from reference)	Required
duration	Integer	Video length in seconds (5 or 10)	5
resolution	String	Video quality (480p or 1080p)	1080p
aspect_ratio	String	Video format (16:9, 9:16, 1:1, etc.)	16:9

🎬 Example Usage

Using Replicate Website

Visit https://replicate.com/rinesh/not-real-konsistent and use the web interface.

Using API

import replicate

output = replicate.run(
    "rinesh/not-real-konsistent:latest",
    input={
        "prompt": "a serene mountain landscape at sunset",
        "duration": 10,
        "resolution": "1080p",
        "aspect_ratio": "16:9"
    }
)

Using cog CLI

# Generate video from text only
cog predict -i prompt="birthday party celebration"

# Use reference image
cog predict -i input_image=@photo.jpg -i prompt="make it look vintage"

# Custom settings
cog predict -i prompt="ocean waves" -i duration=10 -i resolution="1080p" -i aspect_ratio="9:16"

🛠️ Technical Details

Models Used

• FLUX 1.1 Pro: Text-to-image generation
• FLUX Kontext Pro: Image editing with text guidance
• SeeDANCE 1 Pro: Image-to-video animation

Architecture

• Built as a Replicate pipeline model
• Runs on CPU infrastructure
• Automatic model chaining with authentication
• Returns both intermediate image and final video

🏃 Local Development

Prerequisites

• Python 3.11+
• Docker Desktop
• Replicate account and API token

Setup

# Clone the repository
git clone https://github.com/yourusername/not-real-konsistent.git
cd not-real-konsistent

# Install dependencies
pip install -r requirements-local.txt

# Set your Replicate API token
export REPLICATE_API_TOKEN="your_token_here"

# Run local test server
python app.py

Testing with cog

# Test locally with cog
cog predict --use-replicate-token \
  -i prompt="test video" \
  -i duration=5

📦 Deployment

Deploy to Replicate

# Login to Replicate
cog login

# Push as pipeline model
cog push --x-pipeline r8.im/yourusername/your-model-name

Important Notes

• Must deploy as pipeline model (--x-pipeline flag)
• Requires replicate>=1.1.0b1,<2.0.0a1
• CPU-only execution
• Cannot overwrite existing non-pipeline models

📄 License

This project is licensed under the MIT License. See the models' individual licenses:

• FLUX models: Check Black Forest Labs licensing
• SeeDANCE: Check ByteDance licensing

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📞 Support

For issues or questions:

• Create an issue on GitHub
• Contact through Replicate platform
• See CLAUDE.md for development notes