Hybrid prompt builder — type a subject, pick medium / lighting / color / finish dropdowns, 🎲 per field, or one-click 15 presets. Natural-prose output, ready for Replicate / fal.ai / ComfyUI / Freepik.
PRESETS · Quick presets
SubjectFocal subject in plain English — specific beats clever (e.g., 'a young woman in linen shirt looking past camera').
Medium / Style00Pick 1–2 (e.g., fashion editorial + 35mm film)
Mood00Pick 1–2
Camera / Composition
Lighting
Colors00Pick 1–2
Finish00Pick 1–3 (sharp focus, film grain, etc.)
Model variantpro = best quality / dev = open-source / schnell = fastest (4 steps)
HOW TO USE
Type a subject (English, the more specific the better)
Pick medium (photorealistic / editorial illustration / etc.) + lighting + color + finish, or 🎲 randomize all
Pick model variant (Pro = quality, Dev = open-source, Schnell = fast, Ultra = full-size)
Keep prompt ≤ 350 chars — Flux weights front-loaded words most
Copy, then open Replicate, fal.ai, ComfyUI, or Freepik — paste into the prompt field and pick the model you noted
FAQ
Q01Is this generator free? Do I need an account?+
Completely free, runs in your browser (no API calls, nothing stored). No signup — open and use.
Q02How does Flux differ from Midjourney?+
Flux takes natural English prose and doesn't support --flag syntax (no --ar / --v). Its strengths are photorealism, editorial illustration, text-in-image, and brand visuals. Midjourney is stronger on stylized / fantasy / painterly output. The two are complementary.
Q03Which model variant should I pick? Does the prompt change?+
The prompt text is identical — the variant only tells you which model to route it to. Pro = highest quality (Replicate / fal.ai paid), Dev = open-source, runs locally (ComfyUI / Diffusers), Schnell = fastest at 4 steps, Ultra = 4MP full-size output. This field is a reminder of 'which model I recommend for this prompt'; you still pick the model in your actual platform's UI.
Q04Why doesn't the prompt include --ar 16:9 flags?+
Flux doesn't parse that syntax. Aspect ratio, step count, guidance scale — all platform-side controls (Replicate sidebar / ComfyUI node / fal.ai sliders). Writing them in the prompt just adds noise the model treats as description words.
Q05Who makes Flux? How does it relate to Stable Diffusion?+
Flux is a 2024 release from Black Forest Labs — the same team that founded Stability AI and later left to start this company. Technically it's a step beyond SD3, particularly strong at text-in-image (prompting 'DESIGN STUDIO' in the image really renders that text), skin texture, and photorealistic detail.
Q06I see Flux prompts often use 'sharp focus' / 'HDR photography' — should I add those?+
Yes, 1–3 of them (our 'Finish' field is this category). Flux responds strongly to these post-production terms — they tell the model 'I want photo quality, not illustration quality'. For realism: sharp focus + natural skin texture. For editorial feel: subtle film grain. For product: professional retouching + crisp edges.
Q07Why is ≤ 350 chars recommended? Will 500 crash Flux?+
Won't crash — Flux technically accepts 1000+ tokens. But Flux weights prompt openings most; later words fade. Within 350 chars almost everything gets read. Past that, trailing modifiers get ignored — writing more doesn't help. The preview shows a budget bar — amber at 280+, red at 350+.