02 / GENERATE

Type a sentence.
Get an image.
From nothing.

No source photo, no reference, no upload. Describe what you want and the model invents it from scratch — in any style, any aspect ratio, any resolution up to 4K. Make a product shot, a poster, a concept render, or an entire mood board's worth of variations.

Time per batch
~4 seconds
Variations
4 per prompt
Up to
4096 × 4096
Generation result 1 Generation result 2 Generation result 3 Generation result 4
Prompt
Generating…
Possible images
24
Built-in styles
3
Quality models
4K
Max resolution
01 · Anatomy of a prompt

The model speaks English.

No tags, no weights, no negative prompts. Just describe an image the way you'd describe it to a photographer. Five ingredients usually do it.

A weathered fisherman in a yellow raincoat, low-angle portrait, backlit by overcast sky, shot on Kodak Portra 400, grainy film aesthetic, muted color grade, slight vignette

Five ingredients

Mix any combination. Skip what doesn't matter. Order doesn't matter much — the model parses meaning, not syntax.

Subject
What or who is in the image — the noun.
Composition
Framing, angle, distance, what's in shot.
Lighting
Direction, mood, time of day, source.
Style
Photographic medium or illustrative aesthetic.
Modifier
Color grade, texture, post-treatment — the finish.
02 · Style presets

Twenty-four styles.
One click.

If you don't want to write a style brief, pick one. Each preset is a tuned prompt suffix that nudges the model toward a coherent aesthetic.

Photo
Cinematic 35mm
Photo
Portra 400
Photo
Studio commercial
Photo
Editorial fashion
Photo
Documentary
3D
Soft 3D render
3D
Clay claymation
3D
Isometric scene
Illustration
Flat editorial
Illustration
Watercolor sketch
Illustration
Children's book
Illustration
Vintage poster
Art
Oil painting
Art
Impressionist
Art
Pen & ink
Concept
Cyberpunk neon
Concept
Sci-fi matte
Concept
Fantasy painting
03 · Models

Three engines.
One workspace.

Pick speed for thumbnails and exploration, balance for everyday work, or premium for client deliverables.

Sketch

Fast

For exploration and ideation. Lower resolution, smaller variations, generous with concepts. Use when you're brainstorming, not finishing.

~2s
Per image
1024px
Max size
0.5
Credits / image

Studio

Balanced

The default. Strong prompt fidelity, photoreal grade, and crisp detail at HD. Use for most social, web, and product work.

~4s
Per image
2048px
Max size
1
Credit / image

Master

Premium

For finals you're handing to a client. Maximum detail, best typography rendering, large-format ready. Slower, more expensive — and worth it.

~12s
Per image
4096px
Max size
3
Credits / image
04 · Try it here

See it work,
without signing up.

Pick a prompt below to see the kind of image Generate produces.

Pick a prompt

Each preset shows a different model + style combination.

prompt: A weathered fisherman in a yellow raincoat, low-angle portrait, backlit by overcast sky, shot on Kodak Portra 400
Preview Portrait result Product result Landscape result Illustration result
06 · Prompt craft

How to write
a good generation prompt.

The model wants direction, not decoration. Use specific nouns. Skip generic praise. The model knows what "beautiful" means and doesn't need to be told.

Do

Be concrete & specific

  • "A 70s Polaroid of a brown Labrador on a porch"
  • "Product still life, three perfume bottles, marble surface, hard shadows"
  • "Editorial portrait, side light, soft focus, shot on Hasselblad"
  • "Flat illustration, three characters, blue palette, isometric"
Avoid

Overstuffed or empty prompts

  • "Beautiful stunning amazing image" — empty calories
  • "4K HD ultra realistic masterpiece" — model handles quality
  • "((cat:1.4)) ((sunset:1.2))" — no weights here, plain English
  • "Make it good" — describe what good means
07 · FAQ

Questions, answered.

What aspect ratios are supported?

Square (1:1), portrait (2:3, 3:4, 4:5, 9:16), landscape (3:2, 4:3, 5:4, 16:9), and ultrawide (21:9). Custom dimensions on the Master model.

Can I get the same image twice?

If you keep the same prompt and seed, yes. By default seeds are random — for reproducible results, pin a seed in advanced settings.

Can the model render text?

Yes, much better than older models. Short signage and headlines render reliably on Studio and Master. Body copy and small text is best added as a real text layer afterward.

How many variations per prompt?

Four by default. You can set it to 1, 2, 4, or 8. Each variation counts toward credits.

Who owns the generated images?

You do. Anything you generate is yours to use commercially or personally, subject to our Terms. We don't train models on your outputs.

What does it cost?

0.5 to 3 credits per image depending on the model (Sketch, Studio, or Master). Full breakdown on Pricing.

What do you want to make?

Open the workspace. Type a sentence. Watch it appear.