Want an email when I post Day 3?
- Go here: https://higgsfield.ai/image/openai_hazel
- Add your reference image
- Choose "GPT Image 1.5" or "Nano Banana 2"
- Paste this prompt:
imagine this person as a giant, in a tilt shift photograph. Make sure to preserve as accurately as possible the person's facial features. The city appears toy-scale, with realistic model textures: slightly exaggerated paint, crisp edges, tiny imperfections, and handcrafted detail. The street is busy but orderly, full of life.
Tilt-shift photography effect: Extremely shallow depth of field, Sharp focus on the giant and a small plane of the street beneath her. Foreground and background rapidly blur into creamy bokeh. Strong miniature illusion, as if the city is a physical scale model. Optical realism consistent with a real tilt-shift lens, not digital blur
Lighting: Natural shadows beneath the giant and the tiny figures, grounding them physically in the scene.
Color palette: Rich saturation
Camera & realism details: Macro photography feel, High dynamic range, Ultra-sharp textures where in focus, Natural film-like grain, Photorealistic skin, fabric fibers, and materials, Cinematic composition with strong depth cues
Mood & tone:
Whimsical, playful, magical realism. A sense of wonder and delight, as if a normal city moment has been gently bent by imagination rather than fantasy chaos.
Avoid: CGI look, cartoon style, plastic textures, overly smooth skin, exaggerated facial features, unrealistic anatomy, fantasy costumes, surreal distortions, warped buildings, perfect symmetry, artificial lighting, digital illustration style.
5. Hit Generate (I recommend generating 4 images at a time to get options)
6. Once you have your image, go here to create a video: https://higgsfield.ai/create/video and choose Kling O1
7. Add your start frame (your original photograph) and end frame (the giant photograph you generated)
8. Paste this prompt (adjust it to your shot)
the alleyway zooms forward, the buildings drop down and fold in on each other, the woman slowly squats down, while she looks downward
9. Set Duration to 5s and set model to Kling O1
10. Hit Generate
Note: I recommend generating 2-3 videos at the same time, as you'll get different results even with the same settings