The automatic model selection is smart - choosing between 20+ SOTA models manually would be a nightmare. Quick question: how does the AI Director handle visual consistency across scenes? That's usually where AI video generation falls apart - character looks change, lighting shifts randomly, style drifts. Does CrePal have some kind of style/character lock mechanism? Also curious about the 'professional filmmaking logic' - what does that mean practically? Like shot composition rules, pacing decisions, or something else? The real-time preview with feedback sounds useful. Can you iterate on specific scenes without regenerating the whole thing? That would save a ton of compute time.