When using Mochi 1, we collect and process the following types of information:
Text prompts submitted for video generation
Technical metrics related to model performance
System specifications and GPU usage statistics
Generated video outputs for quality assessment
2. Use of Information
We use the collected information to:
Improve the AsymmDiT architecture and model performance
Enhance prompt processing and video generation quality
Optimize system resource usage and efficiency
Research and develop advanced video generation techniques
3. Data Processing
Our 10B parameter model processes data locally on your GPU infrastructure. While prompts and outputs may be temporarily stored for processing, we implement the following safeguards:
Local processing on user-provided hardware
Temporary storage only during active generation
No permanent storage of user prompts or outputs
Optional telemetry for performance optimization
4. Technical Data Security
To protect your information, we employ:
Secure local processing protocols
Encrypted data transmission when applicable
Industry-standard security measures for any cloud components
Regular security audits and updates
5. Research and Development
As an open-source research project, we may use anonymized technical data and performance metrics to advance video generation technology. This includes: