Monday, January 23, 2023

form 2



Dropped back to the Shi-Labs model for the resynthesis for this 'form 2' post (with it's characteristic clipped saturated output).  Both this and the 'form' one are starting with a gallery show output image from the 'baroque' post as input for one shot resynthesis.  The 'form' post is using the new model.  Both are manipulating the latent embedding derived from the image input for the resynthesis.


 

No comments: