Comment on Stable Diffusion forming images from text: image snapshots at each stepparentComments−ShamelessC4yWhat is happening is that the noiser/earlier timesteps are responsible for low-frequency features, while the final timesteps are responsible for high-frequency features.https://dsp.stackexchange.com/questions/1637/what-does-frequ...
Comments
What is happening is that the noiser/earlier timesteps are responsible for low-frequency features, while the final timesteps are responsible for high-frequency features.
https://dsp.stackexchange.com/questions/1637/what-does-frequ...