Z-Image: full sentences, subject first
Z-Image reads full sentences rather than keyword lists, and it's particularly strong with lighting and on-product text — a label comes out legible. Structure the prompt subject first, then lighting.
A neon orange and turquoise aluminum can, label reading WAVE in bold white letters, standing on a white studio table. Soft rim light along the top edge, warm key light from the left.
FLUX.2 Klein: a reference, then word order
FLUX.2 Klein works from a reference image: attach one or more through References in the prompt bar, then describe the scene you want built around them. It's fast and handles complex scenes with several elements at once, and it reads whatever comes first in the prompt as the priority element.
The WAVE can from the reference, resting on a striped pool towel beside sunglasses and a half-peeled orange, poolside, bright afternoon light, shot on 35mm film, shallow depth of field.
Qwen Edit: short, specific edit instructions
Qwen Edit also starts from an existing image, but it's built for precise edits rather than new scenes. It works best when you quote the exact change and specify what should stay untouched.
Keep the WAVE can exactly as it is — same shape, same colors, same label text. Replace the background with a rocky tide pool at sunset, matching the can's current angle and lighting.
FLUX.2 Klein