Every prompt below was generated with GPT Image 2 in Made, and the output next to it is what came back — not a curated stock image. Copy a prompt as a starting point, then use the guide underneath to adapt it.
Original results
These examples use media produced for Made. Prompts are included when the production record allows them.
“Candid portrait of a coffee roaster in her thirties checking beans at a vintage drum roaster, steam and warm tungsten light, charcoal apron. 35mm film still, soft natural light, muted earth tones with one copper accent, shallow depth of field, subtle film grain, candid and unposed, no text, logos, or watermark.”
“Product photo of a matte kraft coffee bag on a walnut counter, printed label reading exactly "SLOW MORNING" with a smaller line "Ethiopia Guji · Light Roast". Clean editorial design, muted earth tones with one cobalt accent, subtle paper texture, sharp legible type, no other words or logos.”
“Hand-lettered chalkboard menu on a white tiled café wall reading exactly "ESPRESSO 3 / CORTADO 4 / OAT LATTE 5", morning light across the tiles. Clean editorial design, muted earth tones with one sage accent, subtle paper texture, sharp legible type, no other words or logos.”
“Overhead pour-over setup on natural linen: gooseneck kettle, ceramic dripper, glass carafe, scale, one cobalt cup. 35mm film still, soft natural light, muted earth tones with one cobalt accent, shallow depth of field, subtle film grain, candid and unposed, no text, logos, or watermark.”
“Wide shot of a small coffee roastery at dawn, burlap sacks stacked by the wall, copper cooling tray, dust floating in light beams. 35mm film still, soft natural light, muted earth tones with one copper accent, shallow depth of field, subtle film grain, candid and unposed, no text, logos, or watermark.”
“Cargo bike loaded with coffee sacks parked on a cobblestone street at golden hour, empty sidewalk, long shadows. 35mm film still, soft natural light, muted earth tones with one mustard accent, shallow depth of field, subtle film grain, candid and unposed, no text, logos, or watermark.”
Describe the subject, composition, materials, lighting, and any exact display text in quotes.
Choose square HD or 4:3 landscape. Made starts at low quality, the cheapest; raise it for the final.
Upload one image and describe the change when you are editing rather than generating.
Inspect typography, hands, and fine detail, then adjust one constraint at a time or switch to high quality for the final.
Iteration is cheap here: GPT Image 2 costs 1–22 credits / image, so rewriting a prompt and rerunning it is a normal part of the workflow, not a budget decision. See the full GPT Image 2 pricing breakdown.
It excels at photorealism, exact prompt adherence, complex multi-element layouts, and high-quality typography — the demanding briefs where cheaper models drift.
More tested prompts across all models live in the prompt library, and the full GPT Image 2 capability rundown is on the model page.