
Loading…
Loading…

Creator · Arina_hoqe
View original source ↗Use the attached face reference image as the ONLY facial identity source. Preserve the exact face, facial proportions, eye shape, nose, lips, jawline, skin tone, hairstyle, and overall identity with extremely high accuracy. Do not change the person's identity. Only replace the face with the referenced face while keeping the pose, clothing, composition, and environment exactly as described. Create an ultra-realistic full-body portrait of the {argument name="subject identity" default="subject crouching on the floor beside a gray-and-white cat wearing a small checkered bandana"}. The subject smiles gently while looking at the cat with a warm, affectionate expression. Behind them is a warm cream-colored wall illuminated by soft golden window light. On the wall, create a large hand-drawn manga illustration of the same girl leaning forward and gently touching the cat's paw with one finger. The manga cat should mirror the real cat, standing on its hind legs while reaching out with one paw. The illustration should seamlessly blend into the wall, creating a magical interaction between reality and manga art. The subject wears {argument name="clothing" default="an oversized charcoal-gray zip hoodie over a white T-shirt, loose blue jeans, and beige-and-black casual sneakers"}. The {argument name="manga style" default="manga drawing features elegant black ink line art, pencil sketch textures, expressive anime eyes, soft cross-hatching, flowing hair, realistic proportions, and detailed shading"}. Soft golden-hour sunlight streaming through a window, realistic shadows, cinematic indoor photography, eye-level composition, 50mm lens, shallow depth of field, photorealistic skin, realistic hair strands, highly detailed wall texture, HDR lighting, masterpiece, editorial lifestyle photography, manga mural illusion, ultra-detailed, premium quality, 8K, extremely realistic.Models & stability
Recommended models
Independent retest evidence
Sep 2, 2026

The requested model is recorded, but the generation receipt did not expose a verifiable underlying model ID.
Aug 9, 2026

The requested model is recorded, but the generation receipt did not expose a verifiable underlying model ID.
Keep exploring

A prompt for creating animation-ready pixel art sprites. It tells the model to produce a 16-frame grid of combat actions such as punches and kicks from a character image reference.
Prompt
このキャラクターの戦闘モーションのスプライトシートを作って。{argument name="サイズ" default="128px"}の簡略化された{argument name="画風" default="pixelart"}、4x4=16コマ、背景は透過。{argument name="アクション内容" default="構え4/パンチ4/キック4/被弾4
A three-layer reference sheet — nine body angles, nine expressions, fourteen fine-grained identity anchors — built to feed an image model a locked character identity, paired with an exhaustive negative-prompt list to stop identity drift.
Prompt
Based on the attached original character image, create a "Character Identity Fixed Reference Chart": landscape orientation 16:10, light gray-white background, professional visual reference board layout. The primary goal is for all panels to

A GPT Image 2 prompt that transforms each uploaded photo into a split-screen design, with the original image on top and a vintage watercolor illustration below.
Prompt
Turn each photo I upload into a separate premium design poster. Do not combine multiple images; output each photo separately. Use an overall 3:4 vertical composition divided into two regions with a strict 1:1 height ratio, each occupying 50
Same model
Woman in Sunglasses Stands in Platform Wind as a Train Rushes PastAI Image · GPT Image 2.5 · 2026-10-28
Blue-Line Couple with Golden Hearts on Cream PaperAI Image · GPT Image 2.5 · 2026-10-27
Hand-Drawn Paper Cover with a Small Scene and Ample Blank SpaceAI Image · GPT Image 2.5 · 2026-10-24
Warmly Lit Portrait of a Black-Haired Woman by a Café WindowAI Image · GPT Image 2.5 · 2026-10-24
Six Flash Portraits of a Young Portuguese WomanAI Image · GPT Image 2.5 · 2026-10-23
Portrait of a Curly-Haired Woman in Champagne Gown and Floral PearlsAI Image · GPT Image 2.5 · 2026-10-21Each detail page keeps the original result, the prompt, and retest evidence in one reading path.