The final speaker in this session at the the ECREA 2026 conference in Brno is Cristina Monzer, with a comparison of TikTok aesthetics in political videos between Norway and Romania. The first question, of course, is how we might do this, in practice – there are many new computational tools emerging for this now, but most of these also remain untested to date.
The use of TikTok for political communication is characterised by personalisation, a focus on everydayness, and uneven patterns between politicians of different genders. Politicians must use visual cues in such videos as identity markers that signal alignment with certain audiences, and blend professional and personal authority. This study examined these patterns by comparing the use of TikTok in the 2025 Norwegian parliamentary election and 2025 Romanian presidential elections; it collected between 300 and 600 videos from these elections.
This content was analysed through a process of (semantic) keyframe extraction, person and face detection, assessment of perceived gender, examination of body emphasis and orientation, and classification of background settings, variously using the tools CLIP, YOLO11, and YOLO-Pose. Such processes are difficult to manage, however, and still produce plenty of misreadings.
There are more women featured in Norwegian than Romanian videos; Norwegian women politicians are also framed more often just in upper body or close-up rather than full-body shots, while in Romania the balance is much more even. Settings also tended to vary, and were often miscategorised as outdoors for Norwegian videos when office settings had windows with a view.
