Benjamin Sommer


When I see great images, it feels like I want to simulate them.

Imagine to synthesize an image stream of natural content, for instance the video posted at, in comparable visual fidelity within temporal interactive constraints.

Do we have to employ a suitable data embedded in some high-dimensional space? What about the SDS problem? The sampling problem introduced by Monte-Carlo approaches due to a dimension-mismatch between existing theories and linear computing devices? Remind the relation of visual image synthesis to spatial material synthesis. Recall the challenge of high-frequency distributions within the nonlinear material and transport space!

How can we achieve both effective and efficient synthesis within the visual and acoustic domain, as seen from a theoretical and artistic perspective?

These research questions constitute the basis of my academic endeavours. I do not stop looking for answers. Noteable breakthroughs will be published here, but only on high standards. Therefore, be patient please.