AI Cinematic Video Prompts — The Exact Recipes

These videos aren't pure text-to-video. They use Seedance 2.0 on Higgsfield in image-to-video mode — which means the prompt is only half of the recipe. The other half is a high-quality photo of yourself as the starting frame. The image sets the subject. The prompt controls everything that happens after — the camera move, the lighting, the emotion, the cinematic style. Without a strong starting image, the model invents a face. With your own photo, it puts you inside the scene. What makes a good starting image: - Shot in good lighting (natural or studio) - Sharp focus on the face and upper body - Neutral or minimal expression — the AI builds the emotion from the prompt - 9:16 portrait orientation ideally, or a clean crop to 9:16 works fine Once you have that, paste the prompt below and the model does the rest. No editing. No post-production. 15 seconds.

Keyword to comment: BLOODSHOT Generated with: Seedance 2.0 · Higgsfield · 15 seconds · 24fps · 9:16 Full prompt: `` Extreme close-up on the subject's eyes — bloodshot, darting left and right in terror. Camera slowly pulls back to reveal the subject is crouched behind a rusted metal door, breathing hard, chest heaving. A single flickering industrial light swings above them. Shadows crawl across the wall. Distant footsteps echo and grow louder. The subject presses their back against the door and squeezes their eyes shut. Cinematic 9:16 vertical frame, David Fincher style, desaturated teal-and-orange grade, heavy film grain, 24fps, shallow depth of field, motion blur on shadow movement. ``

Keyword to comment: PHONE Generated with: Seedance 2.0 · Higgsfield · 15 seconds · 24fps · 9:16 Full prompt: `` Medium close-up: the subject holds a burner phone to their ear in a dark parking garage. Their face is partially lit by the phone screen glow — cold blue light cutting through shadow. They listen, not speaking. Their jaw tightens. Something has gone wrong. ``

Keyword to comment: AWAKEN Generated with: Seedance 2.0 · Higgsfield · 15 seconds · 24fps · 9:16 Full prompt: `` The subject jolts awake in a sterile white room — no windows, one flickering fluorescent light. They look around slowly, disoriented. They look down at their hands as if seeing them for the first time. ``

Keyword to comment: NOLAN Generated with: Seedance 2.0 · Higgsfield · 15 seconds · 24fps · 9:16 Full prompt: `` The subject walks directly toward the camera in slow motion, emerging from thick fog on a rain-soaked cobblestone alleyway at 2AM. Each step sends ripples across puddles reflecting neon signage. Wind whips their coat. They do not break eye contact with the camera. Rain streaks catch the light like broken glass. Camera is low to the ground, tracking backward, slowly rising to eye level. Lens flare from a streetlamp. Motion is fluid, controlled, menacing. 9:16 portrait, 0.5x slow motion, Christopher Nolan cinematography, blue-grey colour palette, film grain, 24fps, bokeh rain, cinematic. ``

After generating dozens of these, here's the structure that consistently produces the most cinematic results with Seedance 2.0: