How the pipeline transforms your words into images. Compare quality tiers and models, browse every image we’ve generated, and learn prompting techniques.
Each tier uses a different model. Draft runs SDXL Lightning for speed, Standard uses FLUX Klein 4B, Premium uses FLUX.2 Dev, Ultra uses Gemini 2.5 Flash, and Ultra+ uses Gemini 3.1 Flash for maximum fidelity. Higher tiers inject quality boosters and style defaults.
Skip the blank canvas. Use img-forge to rapidly generate UI concept art — from lo-fi sketches on draft tier to polished visual mockups on ultra. Part of the Stackbilt ecosystem for shipping faster.
All generations across five tiers and three providers — Cloudflare Workers AI (SDXL, FLUX), and Google Gemini. Filter by tier to compare.
Every prompt passes through a three-stage pure-function pipeline. No AI calls until generation — the enhancement is deterministic and transparent.
The pipeline helps, but a well-written prompt still makes the biggest difference. A few guidelines for getting the most out of each generation.
“a robot” works, but “a weathered industrial robot standing in an abandoned factory” gives the model dramatically more to work with. Nouns and adjectives are your primary tools.
The analyzer detects style keywords like watercolor, photorealistic, anime, oil painting, or 3d render. Including one steers the entire output and prevents the pipeline from injecting a default.
Terms like golden hour, dramatic shadows, close-up, or wide angle are detected by the analyzer. When present, the pipeline skips injecting defaults and trusts your intent.
Draft (SDXL Lightning) — fastest, free. Standard (FLUX Klein) — fast 4-step FLUX. Premium (FLUX.2 Dev) — 25-step generation. Ultra (Gemini 2.5 Flash) — Google’s multimodal model. Ultra+ (Gemini 3.1 Flash) — latest Gemini image generation.