Diffusion model 红起来后这块推太快了
mid journey才要商业化就已经过时了
从openai的dalle-2
https://www.reddit.com/r/dalle2?utm_medium=android_app&utm_source=share
到google的imagen
https://www.reddit.com/r/ImagenAI?utm_medium=android_app&utm_source=share
到stable diffusion
https://www.reddit.com/r/StableDiffusion?utm_medium=android_app&utm_source=share
可以生出到这样子
https://imgur.com/a/XPL5AoX
https://mobile.twitter.com/EMostaque/status/1554011833320837120
老实说imagen的code都有了
https://github.com/lucidrains/imagen-pytorch
你抓danbooru有资源训练大量生成色图都不是问题
conditional下去你想的到的功能都能实现
等勇者去尝试了