Prompt the scene, not the subject
Her face, body and hair come from the character record. Spending your prompt on them at best does nothing and at worst fights the identity you picked her for. The prompt's job is where she is, what she is doing, what the light is like and how the shot is framed.
Four levers that do most of the work
In rough order of impact on whether an image looks intentional or generic.
- Light: window light at dusk, harsh flash, streetlamp, candlelit
- Framing: close portrait, waist up, wide shot from across the room
- Action: reading, laughing mid-sentence, about to leave, looking back
- Setting detail: one specific object beats three vague ones
Why templates are worth using first
There are hundreds of generation templates and they encode the framing and lighting decisions that take practice to get right. Start from one, change one variable, and you learn what each lever does without burning credits on guesses.
Video is a different discipline
A clip needs one simple action, not a sequence. Turning to face the camera works. A three-part narrative does not. Video also takes minutes rather than seconds and costs more credits per render, so it rewards deciding what you want before submitting.