Perfect Memory — AI Semantic Asset Management
AI company specializing in semantic asset management and knowledge engineering for multimedia content transformation.
Overview
Perfect Memory, founded in 2008 in Clermont-Ferrand, France, develops AI-powered solutions for managing and extracting value from multimedia content and knowledge assets. The company combines semantic analysis, knowledge graphs, and machine learning to transform how organizations interact with their multimedia collections.
In October 2025, Perfect Memory expanded into open source AI development with the launch of an AI Memory Proxy system featuring persistent user memory capabilities for LLM applications. This Show HN project uses PostgreSQL and pgvector technology with a single API call architecture that simultaneously generates responses and extracts facts, marking the company's strategic shift toward developer-focused solutions.
The company serves clients across media and broadcasting, cultural heritage institutions, sports organizations, and enterprise knowledge management sectors. Their technology enables organizations to unlock value from content libraries through improved discoverability, enhanced metadata, and intelligent content connections.
Key Features
- Semantic Asset Management: Context-aware organization of multimedia content
- Knowledge Graph Technology: Relationship mapping between content elements
- Automated Metadata Generation: AI-powered extraction of descriptive information
- Media Fingerprinting: Content identification and rights management
- Temporal Navigation: Time-based exploration of audiovisual content
- Entity Recognition: Identification of people, places, objects, and concepts
- Multi-format Media Processing: Video, audio, images, and text handling
- Persistent Memory System: LLM applications with fact extraction and categorization
- API-based Architecture: Integration with existing content workflows
Use Cases
Broadcast Media Archive Management
Transform extensive audiovisual archives into searchable assets supporting production and monetization. Automatic metadata generation describes visual content, spoken words, topics, and personalities without manual tagging. Knowledge graph technology establishes connections between related content across different time periods and programs.
Sports Content Exploitation
Maximize value of sports content libraries through automated identification of key moments, players, and plays. Advanced player recognition tracks individuals throughout footage regardless of camera angle. Play classification automatically categorizes game sequences for rapid content retrieval.
Cultural Heritage Preservation
Digitize and organize multimedia collections from museums, libraries, and cultural institutions. Process historical films, photographs, audio recordings, and manuscripts while respecting cultural domain vocabulary. Create rich networks of cultural knowledge through semantic relationships.
Technical Specifications
| Feature | Specification |
|---|---|
| Deployment Options | Cloud, on-premises, hybrid |
| AI Technologies | Computer vision, NLP, speech recognition, knowledge graphs |
| Media Processing | Video, audio, images, text in multiple formats |
| Memory Architecture | PostgreSQL with pgvector, single API call design |
| Fact Categorization | Preference, personal, skill, goal, opinion, experience |
| Importance Scoring | 0.5-2.0 scale with semantic search |
| Integration Methods | REST APIs, webhooks, direct integrations |
| Language Support | Multi-language processing and interface |
| License | MIT (open source components) |
| Processing Performance | Real-time and batch processing options |
Resources
Company Information
- Website: perfect-memory.com
- Headquarters: Clermont-Ferrand, France