The Akan Cinematic Emotions (AkaCE) dataset is the first multimodal speech emotion recognition (SER) dataset for any African language, specifically developed for Akan, a widely spoken West African language. It includes: 385 dialogues from Akan-language movies; 6,162 utterances from 308 speakers (balanced by gender); rich multimodal data: audio, visual, and text; seven emotion labels (Sadness, Fear, Anger, Surprise, Disgust, Happy, Neutral); word-level prosodic prominence annotations — making it the first prosodically annotated dataset for any African language.