Clipto Raises $15M to Build AI Video Memory

The big picture: Clipto, a startup developing an AI video memory layer, announced $15 million in funding at a $250 million post-money valuation. The round saw participation from leading global investors including HSG (fmr. Sequoia China), GL Ventures, EnvisionX Capital, Palm Drive Capital, Hans Tung, Lu Zhang, and 522 Ventures. The company also launched native Windows and Android apps, alongside a new Model Context Protocol (MCP) designed for secure AI agent access to private media libraries.

Why it matters:

• Unsearchable Media: Despite the ease of video creation, vast amounts of personal footage remain unstructured and unsearchable, effectively locking away valuable information from both users and AI systems.
• Privacy Concerns: Traditional AI video tools often necessitate uploading media to the cloud, introducing privacy risks, bandwidth limitations, and increased operational costs.
• Inefficient Workflows: Manual video scrubbing, cryptic file names, and time-consuming tagging processes hinder efficient information retrieval, making existing video libraries largely inaccessible.

How it works:

• Local AI Processing: Clipto transforms raw video and audio into a searchable memory layer using proprietary on-device AI, ensuring media is indexed without ever being sent to the cloud.
• Comprehensive Indexing: The platform combines video understanding, semantic search, transcription, summaries, and structured video notes to convert disconnected footage into a private, AI-readable memory platform.
• Model Context Protocol (MCP): This new protocol allows AI agents to securely search and interact with users’ private media libraries directly on their devices, maintaining data control and privacy.

The catch: While Clipto’s privacy-first, on-device AI approach offers a compelling differentiator, it enters a competitive market where major tech players like Google and Apple already offer integrated AI features within their photo and video ecosystems. User adoption of a dedicated third-party application against deeply embedded OS features presents a challenge, as does the engineering complexity of scaling proprietary on-device AI solutions across a diverse range of hardware configurations.