Bear Character Edit Screenshot

Goal: Create a clean screenshot-like composition showing an AI image editing variation of a simple bear character, with the smaller original/reference result above and a larger edited result below.

Canvas: Wide horizontal canvas, approximately 1200×960 px, very light gray-white background with lots of empty space. Minimal interface feel, like a chat/image generation workspace.

Layout: Place exactly 2 bear illustration cards on the canvas. Card 1 is small and positioned in the upper-right quadrant, with a rounded white image area. Card 2 is much larger, positioned lower-left/center, also inside a rounded white image area. Between them, slightly below the small top bear, place exactly 1 pale blue rounded chat bubble containing Japanese instruction text. On the large bottom image card, include exactly 2 small translucent gray UI controls: one rounded pill button at the bottom-left labeled 「編集」 and one circular icon button at the bottom-right with a simple upload/share arrow icon.

Text content: The single chat bubble must read exactly: {argument name="edit instruction text" default="頬の横から首にかけてのあたりを、もっと細くしてください。"}

Subject details: The subject is exactly 2 versions of the same cute, hand-drawn, upright bear-like character. The character has off-white fur, thick black sketchy outlines, rounded ears with small inner-ear marks, wide circular eyes with black pupils, a large black oval nose, a tiny straight horizontal mouth, pink circular blush marks on both cheeks, short arms hanging down, stubby legs, and small toe lines. The top bear is smaller and slightly slimmer with a more vertical body. The bottom bear is larger and centered in its card, with a rounded head and body, narrowed neck/cheek area, simple symmetrical pose, and a slightly uncertain blank expression.

Visual style: Minimal doodle mascot illustration, rough black ink outline, subtle paper or canvas texture, soft off-white fill, pale pink cheek circles, no shadows except very faint card/UI softness. The overall image should look like a casual GPT image editing interface screenshot, not a polished poster.

Constraints: Use exactly 2 bear illustrations, exactly 1 Japanese chat bubble, and exactly 2 UI overlay controls. Keep the background mostly empty, preserve the asymmetrical placement, avoid adding extra characters, extra text, watermarks, or decorative elements.