Oct 23, 2026
Black Forest Labs' 5.6B open-weights model runs on 8GB GPUs and places every element with bounding boxes. Three copy-paste layouts.
Speridlabs' Apache-2.0 model makes a 1024px image in about 0.7 seconds. Three prompts built around what Turbo does well, plus its limits.
Oct 21, 2026
OpenMOSS's Apache-2.0 model places every word and object from a JSON layout. Three copy-paste layouts: an infographic, a bilingual poster and a web page.
Oct 20, 2026
Krea's new editing model works in seconds and leaves the rest of your photo alone. Three copy-paste edits: a campaign packshot, a one-object swap and a new-outfit portrait.
Oct 19, 2026
A late-'90s horror still, a sheet-ghost road trip and a Victorian spirit photo, written to keep your face and skip the gore.
Multiverse grid, boxed action figure and a Halloween photo booth, with copy-paste prompts that hold your face steady.
Oct 17, 2026
A three-color illustration prompt and a simple way to check where the eye goes first.
Oct 15, 2026
Three complete original prompts, three generated images and one focused change to try with each.
Oct 14, 2026
Plan three visual beats, control a gentle 2D camera move and verify the exported file with FFmpeg.
