Below you can see the sketch I drew (a bit tricky, using your mouse to draw on in the input field — although you do get an eraser) and the resulting image created by the web tool. The result is a bit scary — both influenced by the input as well as the way the tool works, but it does create a somewhat realistic looking result based on the input.
The fotogenerator is composed by De Kennis van Nu en Nieuws en Co / NPO https://www.dekennisvannu.nl/site/artikel/Help-ons-kunstmatige-intelligentie-testen/9163
This webinterface is made bij Christopher Hesse, based on the work of Philip Isola et al: Image-to-Image Translation with Conditional Adversarial Networks https://arxiv.org/pdf/1611.07004v1.pdf
For a better sense of how similar and yet different the results are from this tool, take a look at the other images generated by today’s participants in the daily create: