Unverified50% confidenceFactExact time
Midjourney V3采用CLIP引导（CLIP-Guided）的生成方式，用OpenAI的CLIP模型评估生成图像与文本的语义匹配度
1
Sources
50%
Confidence
Long-term
Relevance
8/7/2026
First Seen
Sources
Related Claims
UnverifiedLAION团队通过OpenAI的CLIP模型对图像与alt文本的语义相似度进行打分，仅保留相关性较高的图文对76% similarVerifiedCLIP模型能将图片和文字映射到同一向量空间，实现跨模态检索75% similarUnverifiedEarly multimodal models like CLIP primarily focused on image-text matching rather than complex UI understanding74% similarVerifiedCLIP（Contrastive Language-Image Pre-training）通过对比学习将图像和文本映射到共享的语义空间73% similarUnverifiedCLIP由OpenAI与微软联合开发，通过对比学习在4亿张图文对上建立跨模态语义理解72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/700707API
curl https://kongchang.com/api/v1/knowledge/claims/700707MCP
get_claim(id=700707)