AI 5 min read

Emma shares a GPT Image 2 prompt for the “coming out of the sketchbook” effect

@AI_CreatorX publishes a template for ChatGPT on X: spiral notebook with a graphite portrait from which the reference person emerges into the real world, with torn paper and an artist's table.

Verified Editorial standards
Portrait generated with GPT Image 2: person emerging from a graphite sketch in a spiral notebook on an artist's table
Example of the thread: transition from 2D sketch to photorealistic figure emerging from the torn paper. Source: @AI_CreatorX

On June 19, Emma (@AI_CreatorX) shared on X a structured prompt for GPT Image 2 in ChatGPT. The scene combines magic realism and workshop aesthetics: a giant spiral notebook on a dark wooden table, a hyperrealistic graphite portrait on the paper and the reference subject emerging from the two-dimensional drawing with torn paper edges, scattered charcoal pencils and a warm industrial lamp.

AI Prompt · How to do it

View of the result: pencil sketch and three-dimensional figure in artistic studio scene

Result published in the original thread. @AI_CreatorX

Steps

  1. Open chatgpt.com with GPT Image 2.
  2. Upload a clear reference photo of the subject (face and torso visible).
  3. Paste the main prompt in English.
  4. Ask for 4:5 format if the model does not respect it; Adjust pronouns and clothing to the subject of your reference.

Prompt to copy

In the center of the frame, a massive oversized spiral-bound sketchbook stands vertically on a rustic, weathered dark wooden workbench. On the paper is a giant hyper-realistic graphite pencil sketch portrait of the same person from the reference image. In a surreal, mind-bending moment, the thick sketchbook paper is violently torn open from the inside, creating jagged white-edged rips. The person is captured mid-motion stepping out of the two-dimensional drawing into the real three-dimensional world. Their right hand firmly grips the torn paper edge while one foot is planted on the wooden table and the other is still emerging from the sketchbook. Preserve the exact facial features, hairstyle, glasses, beard, smile, and likeness of the reference person with high identity consistency. He is dressed in a premium monochromatic beige outfit: a light beige textured overshirt worn open over a cream crew-neck t-shirt, slim-fit beige trousers, clean white leather sneakers, a brown leather crossbody sling bag across the chest, and a black digital watch on the left wrist. Cinematic dramatic lighting from a single warm industrial overhead lamp casts long shadows and creates a moody high-contrast atmosphere. The surrounding workbench contains scattered charcoal pencils, graphite sticks, sketching tools, a speckled ceramic coffee mug, vintage leather-bound books, and a metal artist palette knife. In the softly blurred background, a corkboard displays paint tubes and artist notes, while a ceramic jar holds assorted paintbrushes. Ultra-detailed textures, photorealistic skin, realistic paper tears, shallow depth of field, volumetric lighting, professional photography, masterpiece quality, high dynamic range, 8K resolution, creative intensity, art-to-reality transition, magical realism.

Negative prompt (optional)

low resolution, blurry face, distorted features, extra fingers, extra limbs, duplicate person, inaccurate likeness, cartoon, anime, CGI look, overexposed, underexposed, bad anatomy, deformed hands, warped glasses, cropped feet, floating objects, unrealistic paper tear, watermark, text, logo, frame, oversaturated colors

We have corrected only pronouns and «person» in the main block for readability; the original mixed "women" and "He." Adjust clothing and features to your reference. Emma also indicated 4:5 ratio, 50mm lens and short depth of field.

What makes this prompt special

The trick is not just "photorealistic portrait": the composition requires a double narrative plane (2D sketch + 3D body), a credible torn paper texture and an artist environment loaded with concrete props (speckled ceramic mug, leather books, palette knife). That anchors the scene and reduces the generic result of “person floating in white studio.”

Emma also packages useful metadata—negative prompt, 4:5 ratio, 50mm lens—in almost JSON format, something common in accounts that optimize results in GPT Image 2. The thread exceeded 580 views, with 38 likes and 20 responses in the first hours.

Tips for better results

Use a reference with good front lighting and without harsh filters. If the model does not tear the paper well, repeat adding "jagged torn paper edges with visible paper thickness." For a darker look, turn up “moody high-contrast”; For a cleaner workshop, reduce the number of objects on the table.

Before publishing

Scenes generated from real photos may involve image rights. Do not use other people's faces without permission. Review OpenAI's policies for synthetic content and commercial use.

Most read

What readers are engaging with right now.