13 | Keep the whole image from blurring. Prompt design for slow-sync style portraits | Airi-chan
Introduction
“Night city, slow shutter, photos with movement.” If you write just that, you might get the atmosphere, but the face might end up blurred as well.
So, let's think about it by dividing it up a little.
Instead of blurring the entire image, first decide on a “static layer,” and make everything outside of that a “motion layer.” In this example, I kept the face and upper body static, while letting the background lights and road reflections flow.
The core of the design is the “static layer” and the “motion layer”
1. Static layer
This is the area the reader looks at first. For a portrait, this includes the face, eyes, upper body, and key points of the outfit. If they are holding a product, include the product here as well.
“Make the subject sharp” is too vague. List the specific parts, such as the face, both eyes, contours, and upper body.
2. Motion layer
This is the area that conveys movement and time. Limit this to background lights, road reflections, distant passersby, or the edges of clothing.
The key here is not to be too greedy. If you blur the hair, face, hands, clothes, background, and buildings all at once, the priority between the subject and the background disappears.
3. The role of light
In this example, I used direct flash to freeze the subject and a slow-shutter style expression to blur the background. In image generation, rather than reproducing strict shooting settings, this is used to clearly define which area is responsible for which visual effect.
Example commentary
I chose a fictional underpass after the rain. Wet road surfaces make it easy to show light trails, and the cyan and amber reflections support the subject's cobalt blue coat.
The subject is slightly to the right of center. I left space in the upper left to make it easier to repurpose for X posts or note headers. I put the hands in the pockets to maintain natural walking posture while avoiding finger distortion.
Final prompt
A pedestrian underpass in a completely fictional city on a night after the rain. A realistic editorial photo of one completely fictional adult woman walking diagonally toward the camera. Short black hair, light cobalt blue coat, charcoal-colored clothes, small plain silver bag. Do not resemble any specific real person, celebrity, influencer, or public photo. Divide the image into a “static layer” and a “motion layer.” Use direct flash to keep the face, both eyes, skin, upper body, coat collar, and bag sharp. In a 1/8 second slow-shutter style, blur the background cyan and amber lights, wet road reflections, two distant anonymous passersby, and only the hem of the coat horizontally. Do not create afterimages or double exposures on the subject's face and torso. Vertical 4:5, eye level, 35mm lens style, medium shot up to the knees, subject slightly to the right of center. Leave space in the upper left. Do not include text, logos, brands, watermarks, identifiable locations, resemblance to real people, additional subjects, duplicate faces or bodies, double exposure of faces, blurred eyes, visible fingers, distorted hands, excessive skin smoothing, or CG/3D expressions.
English prompt
A photorealistic editorial portrait in a completely fictional pedestrian underpass at night after
rain. One fictional adult woman walks diagonally toward the camera, with short black hair, a
cobalt-blue lightweight coat, charcoal clothing, and a plain small silver bag. Divide the image
into a frozen layer and a motion layer. Use direct flash to keep the face, both eyes, skin,
upper body, coat collar, and bag sharply frozen. Simulate a 1/8-second slow shutter only on
cyan and amber background lights, wet pavement reflections, two distant anonymous
pedestrians, and the coat hem, flowing horizontally. No motion trails or double exposure on
the main subject’s face or torso. Vertical 4:5, eye level, 35mm lens, medium knee-up
framing, subject slightly right of center, negative space in the upper-left. No text, logos,
brands, watermark, identifiable location, resemblance to real people, duplicate subject,
double face, blurred eyes, visible fingers, distorted hands, excessive skin retouching, CGI, or
3D rendering.
JSON variable example
{
"location": "Completely fictional urban underpass after rain",
"subject": "One completely fictional adult woman",
"appearance": "Short black hair, cobalt-blue lightweight coat, charcoal clothing, silver bag",
"frozen_zone": "Face, both eyes, skin, upper body, coat collar, bag",
"motion_zone": "Cyan and amber background lights, wet pavement reflections, two distant anonymous pedestrians, coat
hem",
"motion_direction": "Horizontal",
"flash": "Direct flash to freeze the main subject",
"shutter_effect": "1/8-second slow shutter style",
"composition": "Subject slightly right of center, negative space in the upper-left",
"camera": "Eye level, 35mm, knee-up",
"aspect_ratio": "4:5",
"palette": "Cobalt, charcoal, cyan, amber, small amount of magenta",
"negative_prompt": "Text, logos, brands, watermark, identifiable location, resemblance to real people, additional
subjects, duplicate faces or bodies, double exposure, blurred eyes, visible fingers, distorted hands, excessive skin retouching, CGI,
3D"
}
Jinja2 template
A realistic editorial photo of {{ subject }} taken at {{ location }}. Appearance is {{ appearance }}. Divide the
image into a frozen layer and a motion layer. Use {{ flash }} to sharply freeze the {{ frozen_zone }}. Use {{ shutter_effect
}} to flow only the {{ motion_zone }} in the {{ motion_direction }} direction. Do not create motion trails on the main subject's face or torso. Composition is {{ composition }}. {{ camera }}, {{ aspect_ratio }}. Color palette is {{ palette }}. Prohibited items: {{
negative_prompt }}.
Derivations
Fashion: Freeze the main details of the clothing, flow only the hem and background lights
Music: Freeze the face and instrument, flow only the lighting and the audience's arms
Sports: Freeze the face and upper body, flow the background, wheels, and splashes at the feet
16:9: Place the person in the right third, leaving the left side as negative space for headings
Correction prompts
The face is also blurring
Maintain the background light trails and pavement reflections, but please return only the subject's face, both eyes, contours, and upper body to be completely sharp. Please remove motion trails and double exposure from the face.
The movement is weak
Without changing the main subject, please lengthen the light trails of the background lights and wet pavement reflections in the horizontal direction. Do not add any people or objects.
More people are appearing
Please fix the main subject to only one adult woman, and keep the background to within two distant silhouettes whose faces cannot be identified.
The hands are distorted
Please return the hands into the coat pockets, and remove visible fingers, extra hands, or fused fingers. Do not change the face, clothing, or composition.
Summary
For images of people in motion, adding more blur doesn't necessarily make them better.
Decide where to freeze the motion first, and place the motion blur outside of that area. Yes, this is easy to try.
Start with two lines: "freeze the face" and "blur the background light," then add only the necessary elements.
\ Read from the beginning of this series /
\ See the next work here /