Would be interesting to see latency vs quality tradeoffs here. Are they targeting consumer-facing generation speed or prioritizing fidelity for professional workflows?