GPT Image 2 just quietly became the most useful image tool I've tested for real brand work. Not because it's "better" in a benchmark sense — but because it finally does the things that actually matter: it spells correctly, it handles Arabic and English in the same image, it generates working QR codes, and it produces product photography that looks like it was shot in a studio. I tested 5 specific use cases. Here are the exact prompts: Prompt 1 — Storefront signage (text accuracy test) ▎ A warm, sunlit café storefront with a handwritten-style wooden sign that reads exactly "OPEN". Mediterranean architecture, terracotta walls, hanging plants. Photorealistic, golden hour lighting. Prompt 2 — Bilingual menu ▎ An elegant restaurant menu card showing both Arabic and English text side by side. Dark linen background, gold foil accents, clean serif typography. The menu items should be readable in both languages. Luxury Lebanese cuisine aesthetic. Prompt 3 — Product photography ▎ A studio product photograph of a dark amber oud perfume bottle floating on a black marble surface. Dramatic side lighting, smoke wisps rising from the cap, deep shadows, photorealistic 4K quality. Luxury fragrance brand aesthetic. Prompt 4 — QR code poster ▎ A minimalist brand poster with a scannable QR code in the bottom-right corner. Dark charcoal background, gold typography reading "Scan to explore". The QR code must be functional and high-contrast. Clean modern design. Prompt 5 — Brand social post ▎ An Instagram-style brand post for a luxury Lebanese events brand. Dark background, gold and cream color palette, elegant serif Arabic headline centered, with a thin gold border frame. Sophisticated, editorial mood.

I made a premium shoe ad in 20 minutes using only AI. Here's the entire playbook. No camera, no crew, no shoot day. Just a brief, nine AI-generated stills, three Seedance clips, and a quick FFmpeg stitch. The final spot looks like it came out of an agency that bills for weeks. I broke the whole thing down into a 4-phase tutorial and I'm handing you everything: 📎 TUTORIAL.pdf — the full branded walkthrough (cover to outro). Read this first. 📎 deliverables/ folder — the references so you can copy-paste the exact workflow: - BRIEF.md — the 3-shot structure (Reveal → Range → Payoff) - prompts/keyframes.md — 9 Nano Banana prompts (hero + 7 colorways + runner styleframe) - prompts/clip1_sole.txt · clip2_range.txt · clip3_run.txt — the exact Seedance prompts - scripts/seedance_pipeline.py — parallel Seedance runner (just plug in your API key) - scripts/stitch_ad.py — FFmpeg concat + whip-white transitions + tagline + music - CAPTION.md — the caption I posted on IG Two asks: 1. If you ship one using this pipeline, drop it in the comments — I want to see what you made. 2. Hit me with the bottleneck you run into. The next tutorial will answer whatever breaks first for most of you. The structure (Reveal → Range → Payoff) works for any physical product — skincare, watches, bags, electronics, apparel. Swap the subject, keep the three beats. Let's go. — MB

This is the free companion resource for the Claude Opus 4.7 reel on Instagram. 5 prompts I actually use at my agency, with before/after versions and what changed for 4.7. If you try any of these and want feedback, drop your version in the comments below. I read every one. -MB

Prompt Used: A tall hallway closet door suddenly swings open and a young woman tumbles out upside down, her body catching for a split second on the doorframe before she lands on her back on the polished hardwood floor with a soft thud. A few folded shirts and a sweater fall down around her. She lies there for a split second completely disoriented, then pushes herself up on her elbows and immediately notices the camera in front of her. Her expression shifts instantly — from shocked to amused, a slight laugh escaping her lips. She reaches forward naturally and grabs the camera, pulling it slightly closer toward her face as if she's about to start filming herself. The whole sequence feels completely accidental and real — handheld camera energy, slight natural shake on impact, authentic candid lighting. No slow motion. Real time. Ultra-realistic, photorealistic. 9:16 vertical. She looks directly into the camera, still catching her breath, hair slightly messy across her face, and says with a calm, slightly amused tone: "I have absolutely no idea how I got in there. But I'm AI. That was just a hook. Want to recreate this? I'll show you." No music. Ambient room sound only. The delivery is natural, conversational, self-aware — like she knows exactly what just happened and exactly what comes next. Tag @ai.withmb when you post what you make 🔥

Prompt Used: A large black suitcase sitting on the hardwood floor of a minimalist bedroom suddenly unzips from the inside and a young woman pushes the lid open and climbs out, her hair slightly tousled, a single white sock stuck to her shoulder. She steps onto the floor with a slight stumble, looks around for a split second completely disoriented, then immediately notices the camera in front of her. Her expression shifts instantly — from shocked to amused, a slight laugh escaping her lips. She reaches forward naturally and grabs the camera, pulling it slightly closer toward her face as if she's about to start filming herself. The whole sequence feels completely accidental and real — handheld camera energy, slight natural shake on impact, authentic candid lighting. No slow motion. Real time. Ultra-realistic, photorealistic. 9:16 vertical. She looks directly into the camera, still catching her breath, pulling the sock off her shoulder with a small grin, and says with a calm, slightly amused tone: "I'm not gonna explain how I fit in there. I'm AI. That was just a hook. Want to recreate this? I'll show you." No music. Ambient room sound only. The delivery is natural, conversational, self-aware — like she knows exactly what just happened and exactly what comes next. Tag @ai.withmb when you post what you make 🔥