Describe what you want to see, hit Generate, get an image. Image Generator is the catch-all for "I need a picture, fast" — concept art, splash screens, icons, references for the Pixel Artist, mood boards, tile starting points, you name it. Plug in a prompt, pick a model (or several), and the tool fans the request out across them so you can compare.

This is the front door to every other image-shaped tool on the Desktop.

Features

  • A dozen models, side by side. Nano Banana 2, GPT Image 2 (High & Medium), Flux 2 Klein and Pro, Qwen Image 2.0 and 2512, Grok Imagine, Grok Imagine Pro, Kling V3, Seedream 5 Lite, Z-Image Turbo — pick any combination and the same prompt runs on all of them in parallel. Best way to learn each model's personality.
  • Aspect ratios + resolutions. Square for icons, 16:9 for splash art, 9:16 for portrait references, 4K when you need the pixels. Each model exposes only what it actually supports — no guessing.
  • Generate and Edit modes. Pure text-to-image, or drop in a reference image (up to 14 on some models) and the AI edits or remixes from there. Hand the model a sketch, get back a rendered version.
  • Multiple variations per model. Crank up the count to roll 2, 3, or 4 takes of every selected model in one go. Pick the one you like, discard the rest.
  • Prompt templates. Save prompts you keep coming back to — character sheets, environment matte paintings, item icons. Re-use across projects.
  • Live cost chip. The credit price updates as you change models, ratios, and resolutions, so there are zero surprises before you commit.
  • History sidebar. Every batch you've generated stays one click away. Re-open a past run, tweak the prompt, re-roll.
  • Saved to your VFS. Every image lands in your workspace as a real file — drag straight into the Code Editor, Pixel Artist, Background Removal, Extract Sprites, or any other tool that takes images.
  • Multiple windows. Spin up two windows and generate concept art in one while iterating on the boss portrait in the other.

What you walk away with

  • Images that fit your game, generated to your spec, sitting in your workspace.
  • A sense for which AI model nails which style — you'll quickly learn that Flux 2 does photoreal, Z-Image Turbo is cheap and fast, GPT Image 2 nails typography, and so on.
  • A reference library to feed every other tool downstream — Pixel Artist takes the rendered version and pixelates it, Background Removal cuts the subject out, Texture Atlas Creator packs them.

Heads up: generating images uses credits — each model has its own price based on what the provider charges us. The cost chip shows the total before you commit, and edit mode is priced separately from text-to-image because some models are cheaper for one than the other.

The Image Generator is a force multiplier. Once you stop thinking of it as "AI art" and start treating it as a fast first pass that you then polish in the other tools, the whole Desktop suddenly makes a lot more sense.