Brand systems as inputs to models: a brand-voice dataset (rules + preference pairs) and a text-to-video Space.
Text to short video. SDXL frames stitched with FFmpeg.