Forecast time series data using multimodal models
Generate music from text descriptions and optional melodies