The single biggest upgrade to my shots: stop prompting camera moves as words, prompt them as LENS + RIG
'Cinematic camera movement' gives you random slop. What changed everything was describing the physical camera setup:
Instead of 'slow zoom': 'slow dolly push-in on a 35mm lens, shallow depth of field, the background compresses as we approach'. Dolly (camera moves) vs zoom (lens changes) are physically different and the model respects the distinction — a dolly changes perspective, a zoom flattens.
Instead of 'sweeping shot': 'crane up, the camera rises and tilts down to reveal' — name the rig.
'handheld, subtle operator sway, documentary' vs 'locked off on a tripod, perfectly static' — the stability is a creative choice and naming the RIG nails it.
Focal length changes the whole mood. '85mm, compressed, intimate' vs '24mm, wide, you feel the space'. The model learned real lens characteristics.
You're not describing a vibe, you're speccing a camera department. Kling and Runway especially reward this — they clearly trained on labeled cinematography.