潜空间Latent Space
Latent Space
解决什么问题What problem it solves
图像生成的关键机制——模型在压缩后的空间里处理，而不是原始像素。The key mechanism in image generation — the model works in a compressed space, not on raw pixels.
这是什么What it is
先用编码器把图像压缩到低维空间，生成过程在这个空间进行，最后解码回图像。
An encoder first compresses an image into a lower-dimensional space; generation happens there and a decoder restores it.
为什么需要知道Why it matters
它解释了为什么生图模型要「加载 VAE」。它也解释了为什么压缩过程会影响细节。It explains why image models need to load a VAE, and why that compression affects fine detail.
现在能做什么What you can do now
去导航：找现成的工具、平台、评测与学习资源Directory: ready-made tools, platforms, benchmarks and courses ↗
回到名词地图看第2站的其他 35 个词 →Back to the map: the other 35 terms in station 2 →
出处Sources
- 潜扩散模型（LDM）原论文High-Resolution Image Synthesis with Latent Diffusion Models ↗