CLIP

CLIP (Contrastive Language-Image Pre-training) is OpenAI's foundational multimodal model released in 2021 — trained to produce joint embeddings of text and images in the same vector space, enabling cross-modal retrieval (search images by text, search text by images). CLIP was trained on 400 million image-text pairs from the web using a contrastive learning objective, learning to map related image-text pairs to similar vector representations. The model enabled entirely new capabilities: zero-shot image classification, text-driven image search, image-driven text search, and content-based recommendation across modalities. CLIP's embeddings became foundational to many subsequent multimodal systems including Stable Diffusion (which uses CLIP for text conditioning), DALL-E variants, and countless image-search applications. Released under MIT license with weights on Hugging Face. Successor and competitor models include OpenCLIP (community-trained variants on larger datasets), Apple's variant, and Google's SigLIP. AI governance, AI compliance, and AI risk management programs deploy CLIP for multimodal retrieval applications supporting responsible AI through cross-modal search in enterprise AI deployments.

Centralpoint Routes Multimodal Retrieval to CLIP: Oxcyon's Centralpoint AI Governance Platform powers image-text retrieval with CLIP alongside text-only OpenAI, Cohere, Voyage, BGE, and other embedding models. Centralpoint meters every call, keeps prompts and skills on-prem, and embeds multimodal chatbots into your portals via one JavaScript line.


Related Keywords:
CLIP,,