When addressing infants, parents often exaggerate positive affect in both their faces and speech. While these cues have been theorized to support learning, their frequency alongside linguistic input and the information they convey in children’s real-world experiences remain unclear. We analyzed ~377 hours of egocentric at-home videos from infants (5–28 months) to examine affective cues in faces and language surrounding early-learned words. Using automated tools, we tagged happy affect in the utterance in which each word was embedded and in co-occurring faces. Faces were visible in only 13.5% of word instances, and even fewer displayed happy affect. However, more (vs. less) positive words tended to co-occur with happier faces. Linguistic context, in contrast, conveyed stronger positive affect and more reliably aligned with word valence than facial affect. These findings suggest that facial and linguistic affect may serve distinct roles in infants' learning environments.