The integration of Artificial Intelligence (AI) into emotional education represents a paradigm shift from theory to interactive practice. This study presents the results of the permanent EMOCREA sample, developed at the Interactive Center for Science, Art, and Technology (CICAT) at the University of Concepción, under the framework of the FONDECYT project No. 13220040. The research evaluates the effectiveness of AI-based tools personified in three avatars: El Zorro Culpeo (recognition), La Tortuga Tortulina (regulation), and El Mono Caco (expression). Using a mixed methodology with a sample of 172 participants (students and teachers), psychometric instruments (PANAS), technological interaction analysis, and qualitative evaluations were implemented. The quantitative results reveal a statistically significant correlation, indicating that technological interaction significantly displaces negative affect in favor of positive affect. Likewise, the use of multimodal models and pose estimation algorithms for emotional regulation based on the Alba Emoting technique is technically validated. It is concluded that symbolic and technological mediation reduces the cognitive load of social-emotional learning, validating these environments as effective laboratories for human well-being.