For a black-and-white portrait with selective color, name the object that should stay colored and describe the tones you want in the rest of the frame. For a retro flash look, specify a bright subject, a defined cast shadow and a darker background. These CapCut examples compare short style requests with more detailed lighting instructions at 16:9.
Set the Black-and-White Tonal Targets
The starting photograph was generated through the CapCut AI image generator flow in Video Studio's Image mode at Seedream 4.5, 16:9 and 2K: a man in a blue denim jacket on a plain concrete forecourt with a bright red scooter parked beside him. Two saturated objects, one person, nothing else in the frame.
One photograph, three conversions. The second panel is the one-line request; the third and fourth are the same tonal instruction with one color kept and with two.
The short request, “Make this photo black and white,” changed more than color in this example: it added a border, altered the wall texture and produced a much harsher tonal treatment. If you want to preserve the original scene, state that requirement along with the desired conversion.
A more detailed prompt can specify where you want shadow detail, middle grays and bright tones:
Convert this photograph to high-contrast black-and-white film. Redistribute the tones rather than removing the colour: the dark trousers and the shadow under the scooter sit close to black but keep visible texture, the concrete wall and the white t-shirt sit in the bright grays, and his face keeps a full run of middle grays across it. Only a trace of the frame is allowed to reach pure white, and no large area is allowed to clip to solid black. One colour stays in the picture and nothing else: the scooter's red stays exactly as saturated as it is now, and every other part of the frame, including his blue jacket and the colour in his skin, becomes neutral gray. Fine film grain, a slight darkening at the corners.
The chart compares near-black and near-white pixel shares. Its thresholds of 10 and 245 indicate very dark and very bright regions; they do not by themselves establish complete clipping or permanent loss of detail.
The reported near-black share fell from 14.9 to 8.7 percent with the detailed request, while the near-white share rose to 24.8 percent. The result improved one tonal target and missed another. Inspect the trousers, skin and wall separately, then refine the areas that need attention. These examples do not establish a general limit on what prompting can control.
Choose the Object That Keeps Its Color
Name the object explicitly and ask for the rest of the image to become neutral gray. After generation, check the object’s edges, skin and reflective surfaces for unwanted color. The color maps illustrate where saturated areas remain.
The color maps show where saturated pixels remain in the one-color and two-color treatments.
In the reported measurements, 77.4 percent of the colored pixels fall inside the scooter in the one-color version, compared with 44.7 percent in the two-color version. That decrease is expected when the jacket also stays blue. It describes where color is distributed, rather than proving that one treatment is always more visually effective.
A distinctive object with clean edges makes selective color easier to direct and inspect. Choose whether the scooter, coat or another detail should be the focal point, and state that choice in the prompt. If color spills onto nearby surfaces, refine those areas before exporting.
Describe the Retro Flash Lighting
The night look was built from a second generated photograph: a woman on a dark empty pavement in front of a plain brick wall, lit by street light only, with no flash. The one-line request came first here as well.
The night source and its two conversions. The written version put a shadow on the wall with an edge you can point at; the one-line version lit the street instead.
The short flash request brightened both the subject and the street. The sampled face brightness rose from approximately 56 to 93 on the stated 0–255 scale, while the sampled pavement rose from 42 to 146. That result did not achieve the intended bright-subject, dark-background look.
Relight this as a hard on-camera flash snapshot at night. Her face, hands and the front of her jacket are brightly lit from the camera position, with visible detail in the highlights on her forehead and cheekbones. A hard-edged shadow of her shoulder falls onto the nearby brick wall. Make the distant street substantially darker than her face, with only faint environmental detail. A second person passes very close to the camera on the left, blurred and partly outside the frame, while she stays sharp. Add restrained grain, most visible in the dark areas. No text, logos or brand marks.
Mean brightness in sampled face and distant-pavement regions. The detailed request brings the two values closer, but the sampled face is still slightly darker than the pavement.
The detailed version adds a clearly defined cast shadow. Its reported face-to-pavement brightness ratio is approximately 0.96, compared with 0.64 for the short request and 1.34 for the source. The shadow fits the intended style, while the street remains brighter than requested.
For a stronger result, start with a portrait whose background is already subdued, or adjust the background brightness during finishing where the available editing controls allow. Flash illumination weakens with distance, but it has no universal two-meter cutoff. Power, exposure settings, ambient light and subject placement all affect how bright the background appears.
Place the Blurred Foreground Figure
Foreground blur can add a sense of depth to a night portrait. Describe an indistinct shoulder or sleeve entering the edge of the frame, while keeping the main subject sharp.
The example places the passer-by in the distance. To request foreground blur more clearly, describe what the passing figure covers, such as the left edge of the wall. Check that it does not obscure the main subject’s face.
Estimate the Generation Credits
Generation settings and credit estimates recorded on 10 September 2026.
At the displayed rate of one credit per image request, two source images and five conversions total seven credits. This is the estimate for the example requests. Check the current cost shown for your model, resolution and editing mode before generating.
Match the Treatment to the Portrait
Use selective color when a particular object should attract attention, such as a red scooter or yellow coat. Keep one color for a focused accent, or retain a second when the relationship between the objects matters. Check the full composition at its intended display size.
For the illustrated retro flash look, a nearby wall makes the cast shadow easy to see. An open setting can also suit flash styling, but the result will rely more on subject-to-background brightness and the ambient scene. Choose the composition that supports the effect you want.
Use your own portrait or obtain permission from the person shown. Label realistic generated scenes appropriately when sharing them so viewers understand how they were made. Finish by checking skin tones, colored-object edges, shadow placement and background brightness.
Example settings were recorded on 10 September 2026. The source portraits and conversions are AI-generated. The reported measurements describe these examples only; results and available settings can vary.