Image Generator

---

Getting Started

The Image Generator is a powerful, multifaceted tool designed to create a wide array of visual assets for your game. It supports everything from initial concept art and gameplay mockups to specific assets and icons through distinct generation modes.

To start using the Image Generator:

1. Navigate to the Image Generator from the main menu on the left sidebar.
2. Select a Generation Mode: On the left, choose your primary workflow from the dropdown: Generate New, Edit Image, Generate from References, Generate with Style, Edit with Mask, Erase with Mask, or Remove Background. This choice determines which inputs (prompt, base image, additional reference, mask) the form will ask for.
3. Provide Your Input: Depending on the mode, you will enter a text prompt, upload a base image (Image 1), and optionally upload a second reference image (Image 2). When a mode requires a mask, use the Draw Mask button to paint over the area you want to affect. Be as descriptive as possible in your prompts for the best results.
4. Pick an Image Type and Apply Filters (Optional): When the mode supports it, choose an Image Type from the tile picker at the top of the prompt area (Scene Art, Gameplay, Asset Art, Sprite, Game Logo, Parallax Layer, and more). Below the prompt, you can expand Filters to refine Art Style, Perspective, and Aspect Ratio.
5. Click "Generate": Ludo.ai will queue your request and generate three image results by default.
6. Refine, Save, or Download: From the results, click the pencil icon to refine an image further in the Image Editor, click the heart icon to save it to your Favorites, or use the context menu (three dots) to download it. You can also tag your results and search them later — see Organizing and finding your generations.

---

Generation Modes Explained

The Image Generator's workflow is divided into seven modes, accessible from the Generation Mode dropdown on the left. The same seven modes are also available inside the Image Editor when refining an existing image.

Mode 1: Generate New

The most straightforward mode for creating a completely new image from a text prompt and optional filters.

1. Select "Generate New".
2. Choose the Image Type (Scene Art, Gameplay, Asset Art, Game Icon, Sprite, etc.).
3. In the prompt box, describe what you want to create. For example: 4x strategy with zombies.
4. Use the filters to guide the generation. For the example above, you might select Perspective: Top-Down and Art Style: Cel-Shaded.
5. Click Generate.

Mode 2: Edit Image

Modify an existing image using textual instructions. You describe the desired changes in the prompt.

1. Select "Edit Image".
2. Upload your Image 1 — the image that will be edited.
3. (Optional) Image 2: Upload an additional reference image. The AI will use this second image as context for the edits described in your prompt.
4. Provide a text instruction describing the change. For example: turn the astronaut into a zombie or change the background to a night sky.
5. Click Generate.

Mode 3: Generate from References

Create a new image using elements from up to two reference images.

1. Select "Generate from References".
2. Upload Image 1.
3. (Optional) Image 2: Upload a second reference image. The AI will utilize elements from both references when generating the new image.
4. Provide a text prompt to guide how the references should be used or combined.
5. Click Generate.

Mode 4: Generate with Style

Adopt the artistic style of your references for a brand-new image.

1. Select "Generate with Style".
2. Upload Image 1.
3. (Optional) Image 2: Upload a second reference image.
4. Provide a prompt describing the subject you want to create in the style of the provided references.
5. Click Generate.

Mode 5: Edit with Mask

Precise control by editing a specific, masked area of an image.

1. Select "Edit with Mask".
2. Upload your Image 1 — the image to be edited.
3. Click Draw Mask and paint over the area you wish to change.
4. (Optional) Image 2: Upload a second reference image. This can be used to insert a specific object or texture into the masked area, depending on your prompt.
5. Provide a prompt describing the new content for the masked area and click Generate.

Mode 6: Erase with Mask

Removes the masked object completely and intelligently fills in the space.

1. Select "Erase with Mask".
2. Upload your Image 1.
3. Mask the object you want to remove.
4. Click Generate. No text prompt is needed for this mode.

Mode 7: Remove Background

Automatically removes the background from your image. No prompt or mask is needed, and this mode does not consume generation credits.

1. Select "Remove Background".
2. Upload your Image 1.
3. Click Generate.

---

Image Types

The Image Generator supports twenty distinct image types, selectable from the tile picker shown above the prompt. Each type is tuned to produce a specific kind of asset, from broad concept art to sprites, logos, and scrolling layers ready to drop into your game engine.

Picking the right type does a lot of work for you: it steers the composition and framing, sets a sensible default aspect ratio, and — for the types that need it — makes the result seamless or delivers it on a transparent background.

Concept and scene art

• Any type — Non-specified image type, useful when you don't want the model to bias toward any particular category.
• Scene Art — Concept art of scenes or themes. Ideal for visualizing characters, environments, mood boards, and overall art direction.
• Gameplay — In-game screenshot showing gameplay. Use it to visualize how your game will look in action, mock up levels, or test camera angles and perspectives.
• Asset Art — Concept art of a single asset (character, item, prop). The subject is shown in full view against a background that contrasts clearly with it, so it reads as a focused asset study rather than a scene.
• Card Art — Artwork for a collectible or trading card game: a single dramatic subject with a complementary background.
• Splash Screen — Cinematic image for game loading or launch screens, carrying the game's key art and atmosphere.
• Character Portrait — Close-up character portrait for dialogue boxes, menus, or roster screens.

Game-ready assets

These types are delivered on a transparent background, ready to composite directly into your game.

• Sprite — Character or object sprite ready to use in a game engine. Unlike Asset Art, Sprites are generated with flat lighting and shadows removed so they can be animated or composited cleanly.
• VFX Sprite — Isolated visual-effect sprite (muzzle flash, magic impact, explosion) captured at the peak of the effect. Suitable for particle systems or procedural animations.
• UI Element — Buttons, status bars, frames, and other isolated UI elements.
• Item Icon — Small, clear icon representing an in-game item, for inventory, shop, or loot systems.
• Map Tile — Independent grid cell for use in tile-based maps. The tile adapts to the selected perspective — top-down by default, or isometric and high-angle variants with a visible side edge.

Branding and store assets

• Game Icon — App-store icon for a game, always produced at 1:1 regardless of the aspect ratio you select. Use it for store listings and marketing materials.
• Game Logo — A game title logo with stylized lettering, delivered on a transparent background. Put the exact wording you want in your prompt — the title is used verbatim, and it is never invented for you. Leave the wording out and you get a text-free emblem or mark instead.

Backgrounds and scrolling layers

• Background — Static full-screen background for game levels or menus. Focuses on atmosphere, setting, and mood, with no playable characters or UI.
• Repeating Background — Horizontally repeating background for side-scrolling games. Loops seamlessly left to right, and defaults to an ultra-wide canvas so you get a long strip of scenery per generation.
• Vertical Background — Vertically repeating background for vertical scrollers and endless climbers/divers. Loops seamlessly top to bottom, with no ground, horizon, or ceiling to break the illusion, and uniform brightness from top to bottom so the seam never shows.
• Parallax Layer — A single horizontally repeating scene layer on a transparent background, for building parallax scenes one layer at a time. Describe one layer per generation and the layer is anchored appropriately: ground and terrain layers sit flush against the bottom edge, ceiling layers against the top, and floating layers (cloud banks, distant birds, drifting islands) sit in the middle touching neither.

Tiling surfaces

• Texture — Seamless tileable surface for meshes or terrain. Repeats perfectly in both directions across a 3D surface without visible seams.
• Horizontal Tile — Horizontally tileable strip for ground, sky, water, or layered backgrounds. Seamless left to right only.

---

Filters and Options

When the active generation mode supports it, you can refine your request with the Filters panel below the prompt:

• Art Style: Choose the art style you want in the generated image, from a picker of preview tiles covering everything from pixel-art tiers to painterly, 3D, graphic, and craft looks. Pick Any style to leave it to the model. (Not available in the Generate with Style mode, since the style is taken from your reference image.)
• Perspective: Choose the player perspective or camera view (e.g. Top-Down, Side-View, Isometric). Particularly relevant for Gameplay, Map Tile, and Texture images.
• Aspect Ratio: Choose the aspect ratio of the generated images:
  • Auto: Chosen automatically based on the image type — this is the recommended setting.
  • Square: 1:1.
  • Landscape: 4:3, 16:9, or 19:9.
  • Portrait: 3:4, 9:16, or 9:19.

Why Auto is usually the right choice

Each image type carries a default shape that suits it, and Auto applies it for you:

• Sprites, VFX Sprites, UI Elements, Item Icons, Map Tiles, Textures, and Horizontal Tiles default to 1:1.
• Backgrounds, Gameplay, Splash Screens, and Game Logos default to a 16:9 landscape.
• Repeating Background and Parallax Layer default to an ultra-wide 21:9 strip, and Vertical Background to its 9:21 mirror. These extra-long canvases are only reachable through Auto — they are not in the manual list — and they matter, because a scrolling layer is far more useful when a single generation covers a long stretch of scenery.
• Game Icons are always generated at 1:1, regardless of the aspect ratio you select.

Override the aspect ratio when you have a specific target in mind; otherwise leave it on Auto and let the image type decide.

In addition to the filters, the Image Type tile picker above the prompt lets you pick the kind of asset you're generating (Scene Art, Gameplay, Sprite, Game Logo, Parallax Layer, etc.) — see the Image Types section for the full list.

---

The Image Editor: Iterative Refinement

Any image on the Ludo.ai platform can be refined using the powerful Image Editor, which opens in a modal window and contains all the advanced generation modes.

Accessing the Editor: Click the Edit image (pencil) icon on any image thumbnail.

Workflow:

1. The selected image appears in the center as your Base Image.
2. On the right sidebar, you can choose your Edit Mode (Edit Image, Edit with Mask, Generate from References, etc.).
3. Reference 2: Just like in the main generator, you can upload an Additional Reference (optional) in the sidebar to provide extra context or style information for your edit.
4. After you generate a change, the new version appears in the History panel on the right and replaces the image in the center. This image becomes the current/selected image.
5. You can build upon your changes step-by-step.
6. If the generation mode requires a mask, the mask should be drawn directly in the selected image in the center.
7. When finished, select your final desired image from the History and click the Save & Close (checkmark) icon.

This iterative process allows for complex, multi-step image editing, all driven by natural language, dual references, and simple masking.

---

How to Write Effective Prompts

The Image Generator is designed to understand natural, descriptive language. There is no need for special keywords or complex syntax, and there is no hard limit on the length of your prompt. A longer, more detailed prompt gives you more fine-grained control over the final image.

Writing Prompts for New Images

When creating a new image, your goal is to describe the scene as if you were explaining it to an artist. While you can be creative, a good structure often includes:

• Subject: The main focus of the image (e.g., a mystical knight, a futuristic spaceship).
• Action: What the subject is doing (e.g., dueling a shadowy wraith, flying through a nebula).
• Setting & Context: The environment and other details (e.g., amid crumbling ruins, with stars and planets in the background).
• Style & Details: The art style, colors, and specific visual elements (e.g., health bars diminish, with glowing blue cybernetic implants).

Simple vs. Detailed Prompts:

• Simple: green viking
• Detailed: A strong female cyborg with glowing blue cybernetic implants is leaping away from a massive explosion on a busy road of a futuristic city.

Writing Prompts for Modifying Images

When using the Full Edit or Inpaint with Mask modes, your prompts should be instructional. Tell the AI what change you want to make.

• Adding New Elements: Be direct.
  • add a full moon to the sky
  • place a small, glowing lantern on the table
• Changing Existing Elements: Clearly state the element and the desired change.
  • change the car's color to bright red
  • turn the astronaut into a zombie
• Modifying the Overall Image: You can alter the entire scene's mood or style.
  • make the image more dramatic and add a meteor
  • change the style to a Vincent van Gogh painting
  • adjust the color scheme to be warmer with orange and yellow tones

You can also combine multiple instructions in a single prompt for efficiency, such as: add clouds to the sky and make the character's eyes glow red.

---

Limitations and Troubleshooting

The Image Generator is a powerful tool, but it has limitations. Here are common issues and troubleshooting tips:

• Generated images are not relevant to prompts:
  • Be more specific: Add details about characters, setting, colors, and elements.
  • Use filters: Art Style, Perspective, and Aspect Ratio filters guide generation.
  • Pick the right Image Type: Choosing the type that matches your intent (e.g. Sprite vs Asset Art vs Scene Art) has a big impact on the result.
  • Rephrase prompts: Use different keywords to evoke the same concept.
  • Generate variations: Refine relevant results in the Image Editor to get closer to your target.
• AI unable to make requested edits:
  • Simplify instructions: Complex or contradictory edits may confuse the AI. Break them down into smaller steps using the Image Editor.
  • Rephrase editing prompts: Use clearer and more direct language when describing the change.
  • Switch modes: Try a different edit mode — for example, Edit with Mask for localized changes or Generate from References when you want to keep parts of the source image.
  • Start with a different base image: If edits are too complex, a cleaner starting image may yield better results.
• Generated images are not in desired resolution/size:
  • Adjust Aspect Ratio: Use the Aspect Ratio filter to control image dimensions.
  • Pick the matching Image Type: Some types (Game Icon, Background, Texture) imply specific framing or proportions.
• Generation process seems slow:
  • Usage peaks: Generation times may increase during peak hours. Performance should improve after a few minutes.
• Irrelevant or nonsensical results:
  • Rephrase your prompt: Ensure your prompt is clear, concise, and directly related to the image you want to generate.
  • Check for typos: Small errors can lead to unexpected outputs.
  • Start a new generation: Sometimes starting over with a fresh prompt resolves unexpected results.
• Object Removal with Masks not working perfectly:
  • Refine the mask: Ensure the mask fully covers the object and extends slightly into the background.
  • Regenerate: Try generating again after refining the mask.
  • Use Edit with Mask for fine-tuning: After erasing, switch to Edit with Mask with a specific prompt to clean up the area if needed.
• Credit depletion:
  • Check credit balance: Monitor your credit usage in your account settings.
  • Upgrade subscription: If you frequently run out of credits, consider upgrading your subscription plan.
  • Note: Remove Background does not consume generation credits.

If you encounter persistent issues, contact Ludo.ai support or join our Discord server: https://discord.gg/FmTPyugsrR

---

Art Styles

The Art Style filter offers a wide range of distinct visual styles, each with a preview tile so you can see what you're picking. Choose Any style to let the model decide, or pick a specific style to steer the look of every generation.

You can also describe additional style instructions directly in your text prompt — for example, naming a specific artist, era, medium, or color palette ("in the style of Studio Ghibli", "1980s VHS look", "ink wash with red accents"). Prompt-level instructions stack on top of the selected Art Style filter, so use the filter to set the overall look and the prompt to layer on extra detail.

The Image Generator currently supports the following styles:

Drawn / Illustrated

• Cel-Shaded — Flat color areas with hard-edged two-tone shading and bold, even outlines.
• Inked Painterly — Crisp ink linework over painted color, with a vivid palette and strong rim light.
• Illustration — Flat ink illustration with clean ligne claire outlines and flat color fills.
• Western Cartoon — Flat 2D Western cartoon with bold outlines and bright colors.
• Anime — Vibrant modern anime with crisp lines and detailed shading.
• Chibi — Cute, super-deformed characters: oversized head about half the total height, tiny stubby limbs, large eyes, pastel colors.

Pixel Art

These five styles form a ladder from the coarsest, chunkiest pixels to the finest. The first four snap the result to a strict pixel grid, so the pixel size is consistent across the whole image — pick the tier that matches the resolution you're targeting.

• Micro Pixel — The coarsest tier. Large chunky pixels, hard edges, and around sixteen flat colors, in the spirit of the NES / Game Boy era.
• 8-Bit — Clearly visible square pixels with a flat, limited palette, in the 16-bit console tradition.
• 16-Bit — Finer pixels with crisp aliased edges and a broader flat palette.
• 32-Bit — Modern hi-bit pixel art: tiny, densely packed pixels with intricate detail and hard aliased edges.
• Free Pixel — Fine-grained pixel art in the style of 90s console games with a limited flat palette, but without the strict pixel grid of the tiers above. Use it when you want the pixel-art look with more freedom in the detail.

Painterly and traditional media

• Hand-Painted — Hand-painted look with soft visible brushstrokes and color blotches.
• Digital Painting — Airbrushed digital painting with glossy highlights, cinematic grading, and sharp detail.
• Comic Book — American comic ink and color: bold brush outlines, flat spot colors, halftone shading.
• Block Print — Linocut print: carved gouge lines, heavy black-on-white contrast, one or two flat ink colors.
• Sketch — Monochrome graphite pencil line work with crosshatching and an unfinished, hand-drawn feel.
• Watercolor — Translucent layered washes on cold-press paper, with soft bleeding edges and visible paper grain.

3D

• Stylized 3D — Chunky mobile-game 3D with exaggerated proportions.
• Pixar Style — The 3D animated feature-film look: appealing characters and cinematic lighting.
• Low Poly — Low-poly 3D with large flat facets, faceted shading, and untextured solid colors.
• Photorealistic 3D — Lifelike 3D visuals with high detail and realistic lighting.
• Voxel Art — Blocky, 3D pixel-like visuals built from cubic voxels.
• Retro 3D — Mid-1990s PlayStation-era graphics: angular texture-mapped models, blurry low-res textures, jagged edges.

Graphic / Design

• Flat Design — Simple geometric shapes in solid unshaded colors, with no outlines, gradients, or shadows.
• Minimalist — Simple and functional, with minimal detail and lots of negative space.
• Silhouette — Solid black subject shapes with no interior detail, against a single bright flat background.

Atmospheric

• Noir — Black and white with crushed blacks, a hard key light, long shadows, and heavy haze.
• Neon — Glowing magenta and cyan tube-light outlines against near-black, with strong bloom.
• Glitch Art — 2D digital distortion with corrupted pixels, chromatic aberration, and color channel shifts.

Craft / Material

• Claymation — Plasticine models with fingerprints and tool marks, soft rounded edges, and matte clay surfaces.
• Paper Craft — Cut paper shapes stacked in parallel planes, with visible paper thickness and drop shadows.
• Textile — Knitted wool yarn with chunky visible stitches, soft fiber texture, and felted seams.

---

Organizing and finding your generations

Every image you generate stays in your history below the form, and that history grows quickly. The search bar above the results is there to keep it manageable.

Searching

Click the search bar above the results grid and start typing to filter your image history. As you type, the panel suggests matching tags from your library, so you can complete a term instead of remembering its exact spelling. Press Enter or click a suggestion to run the search.

Searches you submit explicitly are remembered: click into an empty search bar and the panel shows your recent searches, ready to re-run with one click.

If nothing matches, the results area tells you so directly rather than showing an empty grid.

Tagging your images

Each image in the history can carry your own tags — a project name, a character, an art direction, a client, whatever helps you find it again. Add or edit tags directly on the image in the results grid; existing tags are suggested as you type, so your vocabulary stays consistent instead of fragmenting into near-duplicates.

Click any tag shown on an image to add it to the search query. Clicking a second tag narrows the search further — tags combine, so you can drill down from "all backgrounds" to "the forest backgrounds for the platformer" in two clicks.

Filtering by date

The search bar also carries a Date range filter. Set a From and To date to restrict the history to images created within that window — useful when you remember roughly when you made something but not what you called it. The range is inclusive on both ends, and it resets when you reload the page so a forgotten filter never silently hides your work.

Paging through results

The history is paginated below the grid. Use the page controls to move through it, and the page-size selector to choose how many images you want to see at once. Your preferred page size is remembered for next time.