IP
Underwater Dreamy Close-up Portrait preview image
Primary reference image

Underwater Dreamy Close-up Portrait

A detailed style-transfer prompt that uses an uploaded photo for identity while creating an underwater-style, cool-toned cinematic portrait.

This is a GPT Image 2 prompt case for Portrait & Character. Use the copy-ready prompt below to generate similar visuals, and review Awesome GPT Image 2 Prompts attribution plus commercial-use rights before reuse.

Need the full prompt set? Use the Portrait & Character topic hub for more related examples, or open the GPT Image 2 prompt library for the full example index, reusable structures, and source attribution.

Free 1K preview

Generate one preview image instantly, no sign-up required.

Free text preview doesn’t preserve reference-image consistency.

2263/4000
Continue on Dovoo

Prompt

Copy-ready prompt

Use the uploaded portrait only as the identity reference. Preserve the person’s real facial features, face shape, skin tone, eye shape, lips, nose structure, and recognizable hairstyle. Do not copy the original clothing, background, lighting, pose, or image quality from the uploaded portrait.

Create a realistic vertical close-up portrait photo of the person’s face captured at a tilted sideways angle, as if the camera is very close to the face through clear water or wet glass. The face fills most of the frame, with one eye closer to the camera, soft calm expression, slightly parted lips, natural gaze looking toward the lens. The head is gently rotated, creating an intimate cinematic composition without unnatural neck twisting.

Style the hair as {argument name="hair style" default="wet, dark, and slightly messy"}, with fine strands floating across the forehead and around the eyes. Add tiny water droplets, soft moisture on the skin, subtle floating particles, and a dreamy underwater-like atmosphere. The skin should look pale, dewy, and realistic, with visible natural texture, soft highlights, and no plastic smoothing.

Use {argument name="lighting" default="cool blue-gray lighting"}, soft diffused highlights, shallow depth of field, cinematic blur, gentle film grain, realistic lens softness, high-detail facial realism, emotional quiet mood, delicate beauty editorial photography. Choose a {argument name="outfit style" default="simple elegant outfit"} that best suits the person’s appearance, such as a minimal light-toned sleeveless top or soft neutral fabric, barely visible at the edge of the frame. No text, no logo, no watermark.

Negative Prompt:

watermark, logo, text, signature, AI-generated text, Show every face clearly with natural, detailed facial features, distorted face, changed identity, wrong facial features, unnatural neck twist, broken anatomy, plastic skin, over-smoothed skin, doll-like face, anime style, cartoon style, heavy makeup, harsh beauty filter, extra eyes, deformed eyes, asymmetrical eyes, bad pupils, messy teeth, overexposed face, low resolution, noisy artifacts, duplicate face, cropped facial features, unnatural hair, dry hair, stiff pose, unrealistic water, random clothing, reference image watermark.

Prompt variables

Editable argument placeholders found in the prompt, with their default values.

3
Variable
hair style
Default
wet, dark, and slightly messy
Variable
lighting
Default
cool blue-gray lighting
Variable
outfit style
Default
simple elegant outfit

Best for

  • - Portrait & Character visual exploration
  • - Image generation and reference-image edits
  • - Reusable briefs that keep source attribution visible

Change these parts

  • - Subject, product, character, or scene
  • - Aspect ratio, camera, lighting, and background
  • - Brand, text, color, and output constraints

Recommended model and settings

  • - Model: gpt-image-2
  • - Input mode: Image to Image
  • - Aspect ratio: source
  • - Commercial status: review original source

How to use this prompt

  1. 1. Copy the prompt and preserve its structure for the first test.
  2. 2. Replace the subject, context, and publishing channel.
  3. 3. Change one camera, lighting, or style variable per iteration.
  4. 4. Inspect the output, source, text, and rights before reuse.

Limitations, source, and reuse cautions

  • - Generated output is an editable draft, not factual, legal, or rights evidence.
  • - Review the linked source, people, brands, logos, text, and third-party media before commercial use.
Imported from Awesome GPT Image 2 Prompts. Attribution is required. Commercial-use status is allowed; review source rights before paid usage.

More cases in this category

Prioritized by category, input mode compatibility, quality, and lower risk.

6

Reuse and source notes

Use this prompt safely after previewing the case.

  1. 1.Copy the prompt or open it directly in Dovoo with the generation button.
  2. 2.Adjust variables, aspect ratio, and reference images for your own use case.
  3. 3.Before publishing or paid usage, verify source rights, attribution requirements, and brand or likeness risks.

Can I use this prompt commercially?

Commercial-use status is allowed. Review the original source, license, brand constraints, and legal requirements before paid usage.

Where does this case come from?

This case is imported from Awesome GPT Image 2 Prompts; keep attribution visible and check the source URL before reuse.