In the area of computer vision, deep learning techniques haverecently been used to predict whether urban scenes are likely tobe considered beautiful: it turns out that these techniques areable to make accurate predictions. Yet they fall short when itcomes to generating actionable insights for urban design. Tosupport urban interventions, one needs to go beyondpredictingbeauty, and tackle the challenge ofrecreatingbeauty. Unfortunately, deep learning techniques have notbeen designed with that challenge in mind. Given their‘black-box nature’, these models cannot be directly used toexplain why a particular urban scene is deemed to bebeautiful. To partly fix that, we propose a deep learningframework (which we name FaceLift1) that is able to bothbeautifyexisting urban scenes (Google Street Views) andexplainwhich urban elements make those transformed scenesbeautiful. To quantitatively evaluate our framework, wecannot resort to any existing metric (as the research problemat hand has never been tackled before) and need to formulatenew ones. These new metrics should ideally capture thepresence (or absence) of elements that make urban spacesgreat. Upon a review of the urban planning literature, weidentifyfivemain metrics: walkability, green spaces,openness, landmarks and visual complexity. We find that,across all the five metrics, the beautified scenes meet theexpectations set by the literature on what great spaces tendto be made of. This result is further confirmed by a 20-participant expert survey in which FaceLift has been foundto be effective in promoting citizen participation. All thissuggests that, in the future, as our framework’s componentsare further researched and become better and moresophisticated, it is not hard to imagine technologies that willbe able to accurately and efficiently support architects andplanners in the design of the spaces we intuitively love.