AI-assisted multimodal analysis of digital media-enabled intergenerational transmission of regional musical intangible cultural heritage: evidence from blind ballad artists' narrative singing in Jindongnan and the Taihang mountains

Digital media has expanded the dissemination of regional musical intangible cultural heritage (ICH), but online visibility does not necessarily lead to intergenerational transmission. This study proposes an AI-assisted multimodal framework to examine the digital dissemination and transmission pathways of blind ballad artists' narrative singing traditions in Jindongnan and the Taihang Mountains. The dataset comprises 2,136 online videos, 184,572 interaction records, 412 youth questionnaires, and 36 field interviews. To avoid target leakage, normalized 30-day dissemination effectiveness was predicted using only content-intrinsic features, creator metadata, and 24-h early engagement indicators. The Heritage-aware Explainable Dissemination Framework integrates textual, visual, audio, platform, and field-informed cultural features. Across five creator-level splits, it achieved an MAE of 0.412 ± 0.014, an RMSE of 0.552 ± 0.018, and an R 2 of 0.706 ± 0.020. SHAP analysis identified early engagement, cultural embeddedness, youth-oriented interpretation, inheritor visibility, subtitles, and village scenes as key predictors. PLS-SEM further identified a significant statistical indirect pathway linking digital exposure with transmission intention via emotional resonance, cultural identity, and learning participation. Given the cross-sectional questionnaire design, these relationships are interpreted as associations rather than temporal or causal effects. The findings indicate that stronger transmission-oriented digital engagement is associated with culturally embedded representation, interpretive accessibility, and participatory learning.