AI 4 min read

Chesny makes a ChatGPT prompt viral for low-key portraits in profile with tubular light

Upload scene and character: the prompt maintains a horizontal light bar, smoke and editorial silhouette. +43K views on X.

Verified Editorial standards
Low-key portrait in profile with white light bar and smoke, generated with Chesny's prompt
Prompt result: same low-key scene with a different character inserted in the silhouette. Source: @chesny on X

On June 17, Chesny published on "This ChatGPT prompt is crazy," he wrote in his June 17 thread: "You just have to copy it, describe your idea and it generates an image exactly of what you have in mind." The post exceeded 43,000 views, 281 likes and 328 saved. In the thread he attaches two images—original scene and result—and the full text in an answer.

AI Prompt · How to do it

Reference scene: profile portrait with tubular light bar and smoke Result: same low-key frame with another character in the silhouette

Reference scene (above) and result with replaced character (below). @chesny

Steps

  1. Open chatgpt.com with image generation active.
  2. Upload the scene reference image (frame, pose and tubular light you want to keep).
  3. Upload the photo of the target character whose facial identity you want to insert.
  4. Paste the prompt below and generate the image.
  5. If necessary, ask for more smoke or more contrast in the outline in a second message.

Prompt to copy

Create a hyper-realistic cinematic low-key portrait based on the uploaded reference image. Keep everything exactly the same: vertical framing, full-body-to-knees crop, subject placement in the center, side-profile pose, dark black background, strong horizontal white tube light behind the subject, smoke/fog around the backlight, subtle lens flare, deep shadows, minimal composition, high contrast, and mysterious editorial atmosphere. Replace only the original person with the character from the newly uploaded target reference image. Preserve the target character's real facial identity, face shape, hairstyle, hairline, forehead, nose, lips, jawline, chin, skin tone adapted naturally to the low-key lighting, and any facial hair if present. The target character must remain recognizable while still appearing mostly in silhouette. The target character must keep the exact same pose and position: standing upright in profile view, facing to the right side, shoulders relaxed, head level, body still, hands naturally down or in pockets depending on the silhouette, with a calm, serious, elegant expression. The side profile of the face must remain clean and clearly outlined against the bright backlight. Change only the outfit according to the target character, choosing the most suitable elegant dark outfit for this scene. Prefer a black or very dark tailored blazer, suit jacket, long coat, or minimal formal fashion piece that preserves the same refined silhouette. The outfit must remain subtle and mostly dark, without distracting details, so the silhouette effect remains dominant. Maintain the exact same lighting setup: one bright horizontal light bar running behind the subject at upper torso level, creating a rim-lit side silhouette; soft smoke or mist spreading around the light; extremely dark surroundings; very limited front lighting; and a faint cinematic lens flare near the lower chest area. Keep the smoke density, backlight intensity, framing, and dramatic mood unchanged. Do not change the background, smoke, light bar position, light intensity, camera angle, crop, pose, subject placement, color palette, contrast, or atmosphere. Only replace the person and adapt the clothing to the target character.

Requires ChatGPT with image generation. The prompt asks for two references: scene + character. Do not use photos of third parties without permission or public figures for commercial purposes without authorization.

What makes this prompt special

Unlike Chesny's studio prompt from June 15 - which describes lighting from scratch - this one from June 17 works as a scene template: fixed vertical framing, profile pose looking to the right, black background, horizontal tubular light bar at torso height, smoke around the backlight, subtle lens flare and mysterious editorial atmosphere. Then he orders the original subject to be replaced only by the character in the second photo, preserving a recognizable facial identity within an almost total silhouette.

The key is in the negative restrictions at the end: "Do not change the background, smoke, light bar position..." - they prevent ChatGPT from reinterpreting the scene and push towards a controlled cinematic face swap. The outfit adapts to the character but must remain dark so as not to break the low-key effect.

How to customize it

If you don't have your own reference scene, first generate a profile portrait with light bar and smoke—or use Chesny's image as a template—and then upload your photo as a target. To change the profile direction, replace “facing to the right side” with “facing to the left side” in the prompt. If the face is too dark, ask: "slightly increase rim light on the profile edge."

Reaction in X

Chesny posted the visual example and text in two consecutive posts on June 17 at 17:14 UTC. The community saved it as a template for covers, posters and promotional material with a luxury campaign aesthetic. The creator regularly shares AI prompts in Spanish and English and is a member of AISC Work.

Most read

What readers are engaging with right now.