From initial sketch to final rendering, realistic models of architectural concepts traditionally take days (if not weeks) to produce. The fantastical facades pictured here — structures sheathed in swaths of multicoloured fabric, glazing encased in Antoni Gaudí–inspired curves — are different. They were created by artificial intelligence.

Using text-to-image A.I. tools (Midjourney is one of the most popular), designers and architects are conceptualizing boundary-pushing envelopes using little more than a string of text. Enter a few words and these tools get to work, scouring image “memories” developed during the A.I.’s neural network training. Seconds later, they generate new forms and invite the human on the other side of the screen to refine, remix or upscale their appearance. A welcome disruption to a manual process or a threat to the human element of creation? Two designers weigh in.