Conventional wisdom holds that populist politicians are more negative, emotional, and angry speakers than others (e.g., Nai 2021). Yet, fringe political figures are also known to act strategically when it comes to their leaders’ images, policy positions, and the language they use to advance their political causes. This pressure may make populist politicians moderate their stronger or more negative emotions in front of a general audience, as in a nationally televised election debate, but to be more expressive in front of supporters. In one of the largest multimedia inquiries into politicians’ emotional expression in different forums (televised debates and interviews; campaign events; the parliamentary setting; freeform web content), we assess whether fringe figures are more moderate in emotionality in front of a larger audience than in front of their core supporters, and how populists compare to non-populist speakers. We use deep learning tools (MEXCA), and/or direct vocal measurements (amplitude and pitch changes), to assess politicians’ emotional activation levels across different settings. As previous research indicates that non-populists engage in emotionally activated speech at similar rates to populists in televised debates (Moez 2024), we investigate whether the forum of speech, and its strategic incentives, drive this result.