Kling, Seedance, Flux, Nano Banana Pro, After Effects, Premiere Pro
This is the mock commerical fully generate by AI.
Step 1: Generate the main character and outfit by midjourney
Prompt:
East Asian woman, real and natural beauty cosmetic model portrait long straight black hair, natural makeup with subtle rosy blush and glossy gradient lips, clear and luminous skin with visible fine pores and natural dewy glow, light skin tone and soft facial features, minimal expression, wearing white spaghetti strap top, clean studio environment, vertical crop from shoulders up, hyperrealistic editorial beauty photography, cinematic detail and tonal depth, Canon EF 85mm lens, f/2.0 aperture for shallow depth of field, soft and even front studio lighting emphasizing real skin texture, ultra-high resolution
Step 2: Tweaking the character over a few iterations to polish it up.
To achieve a higher level of realism for the character, I implemented the JSON prompt for the Nano Banana Pro.
Json Prompt:
Add the following details to this image while keeping the subject's appearance, outfit, and hairstyle unchanged:
{ "camera": { "model": "Mamiya RZ67", "lens": "110mm f/2.8" }, "film": { "type": "Kodak
Portra 800" }, "skin_details": { "pores": "visible fine pores", "texture": "natural micro-relief skin
texture", "vellus_hair": "subtle", "freckles": "natural" }, "shading": { "subsurface_scattering":
"melanin-based", "roughness": { "cheeks": "slightly higher", "t_zone": "slightly lower" },
"specular": { "model": "GGX", "ior": 1.48 } }, "geometry_detail": { "micro_displacement": true,
"normal_map": "high-resolution" }, "lighting": { "key_light": "soft 45-degree", "fill_light": "low
intensity" }, "post_processing": { "plastic_skin": false, "over_retouching": false,
"noise_reduction": "minimal" } }
Step 3: Create the product image.
Step 4: Lock the character:
Utilizing ComfyUI with Flux 2 to lock in the character design.
Step 5: Created a comprehensive shot list using Midjourney and Nano Banana Pro, establishing strict start and end frames for every shot to maintain video stability.
Step 6: To ensure facial consistency across the shot list, I adjust and lock the character's face using the face swap workflow in ComfyUI.
Step 7: Utilized Kling 3.0 and Seedance 2.0 to generate the final video outputs.