Unverified50% confidenceFactExact time
潜在扩散模型将图像生成从像素空间转移到压缩后的潜在空间进行，包含编码器压缩图像、U-Net在潜在空间执行迭代去噪、解码器还原为像素图像三个关键组件
1
Sources
50%
Confidence
Long-term
Relevance
8/27/2026
First Seen
Sources
Related Entities
Related Claims
Unverified在Midjourney所使用的扩散模型架构中，图像先在压缩的低维潜空间中完成逐步去噪，再通过解码器还原为完整图像80% similarUnverifiedModern AI video generation models use variational autoencoders to compress high-dimensional pixel images into low-dimensional latent vector representations, finding transition paths in the latent space rather than performing pixel-level linear interpolation.75% similarUnverifiedImage generation models typically use CLIP or similar cross-modal encoders to map text into high-dimensional vectors, which are then used by U-Net or DiT architectures to progressively denoise and generate images in latent space74% similarUnverifiedImage2 migrated image generation from standalone diffusion models (such as the cascaded diffusion architecture used by DALL·E 3) to an autoregressive token generation paradigm.72% similarUnverified基于扩散模型的数据增强利用Stable Diffusion等生成模型将合成渲染图像风格化为更接近真实照片的外观，同时保留原有几何标注信息71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/807864API
curl https://kongchang.com/api/v1/knowledge/claims/807864MCP
get_claim(id=807864)