Source clip + prompt — 2–15s with audio
Recommended: R2V for multi-ref identity · I2V = one start frame (same template style)