When game studios, indie developers, and concept artists need production-ready reference art at speed, they reach for Leonardo AI. While Midjourney produces stunning gallery-quality images and Flux excels at photorealism, Leonardo occupies a different niche entirely: it's built around the practical needs of game production pipelines — character sheets, environment concepts, prop designs, and UI art.
The platform's model selection, Image Guidance system, Alchemy rendering mode, and deep customization options give professionals a level of control that purely consumer-focused tools don't offer. This guide covers how to use all of it effectively in 2026.
Why Game Artists Choose Leonardo AI
The choice to use Leonardo over alternatives usually comes down to a few specific features that no other platform combines as well:
- Model selection per project: Switch between photorealistic, stylized, anime, and painterly models from a single interface
- Image Guidance (ControlNet): Use reference images to control pose, composition, depth, and character consistency
- Character consistency: Style Lock and Character Reference features help maintain consistent character appearance across multiple generations
- Production-oriented output: Transparent backgrounds, upscaling, and canvas tools integrated directly
- API access: Build Leonardo into custom production tools and pipelines
- Free tier with daily tokens: Accessible for indie developers watching budgets
None of these features is unique to Leonardo individually — but having all of them in one coherent platform is what makes it the professional's choice.
Leonardo's Model Ecosystem
Choosing the right model is the most important decision you make before writing a single word of your prompt. Leonardo hosts dozens of community fine-tuned models alongside its first-party offerings.
Phoenix (First-Party Flagship)
Leonardo's own model, Phoenix is their best general-purpose generator. It produces highly coherent images with strong understanding of compositional instructions. Phoenix is best for:
- Concept art that requires strong composition guidance
- Cinematic illustrations and key art
- Mixed reality/fantasy scenes requiring believability
AlbedoBase XL
One of the most popular community models. AlbedoBase XL specializes in realistic, high-detail renders with strong lighting. It's particularly good for:
- Environmental concept art (both exterior and interior)
- Architecture and structural design reference
- Photorealistic character portraits and close-ups
- Props and equipment with material detail
DiffusionXL (SDXL Foundation)
The base SDXL model gives you broad capability with maximum flexibility. Use DiffusionXL when:
- You want clean output to use as an LoRA training base
- You're doing broad style exploration before committing to a fine-tuned model
- You need maximum compatibility with third-party ControlNet presets
Anime Pastel Dream
Specializes in anime and manga-adjacent illustration styles. Works well for:
- JRPG character designs
- Anime-style portraits and expressions
- Stylized chibi and SD (super-deformed) character concepts
Other Notable Models
- RPG v5: Tabletop and CRPG aesthetic, painterly and detailed
- Dreamshaper XL: High fantasy and sci-fi with painterly quality
- Crystal Clear XL: Crisp photorealistic output with sharp detail
- Absolute Reality: Hyper-realistic human portraits
Prompt Structure for Leonardo
Leonardo uses Stable Diffusion architecture (SDXL and derivatives) for most models, which means prompt structure follows SD conventions rather than Midjourney's natural language style. The most effective structure:
[Subject description], [environment/setting], [lighting], [style/aesthetic], [quality tokens], [medium/rendering]
Example:
female warrior character, full body, detailed armor with crystal inlays, standing in ancient ruins at dusk, dramatic backlit silhouette, fantasy concept art, digital painting, highly detailed, 8k resolution, artstation quality
Key differences from Midjourney prompting:
- Quality tokens matter: phrases like "highly detailed", "artstation quality", "masterpiece", "8k" signal to the model to increase detail density
- Comma-separated tags work well — full sentences are fine but not required
- Negative prompts are essential and should be set separately (see below)
- Model-specific trigger words exist — check each model's documentation for optional activation tokens
Alchemy Mode: What It Does and When to Use It
Alchemy is Leonardo's enhanced rendering pipeline. When enabled, it applies additional processing passes that improve:
- Coherence and anatomy (fewer deformed limbs, better hands)
- Prompt adherence (the final image matches your description more precisely)
- Detail density (more texture, more micro-detail in surfaces)
- Color grading and lighting quality
Alchemy costs more tokens than standard generation. Use it when:
- You've settled on a concept and want higher quality renders
- Anatomy or coherence issues keep appearing in standard mode
- You're generating final deliverables rather than quick explorations
For rapid ideation and exploring many concepts, turn Alchemy off to conserve tokens. Turn it on when you've narrowed down to a direction you want to refine.
Image Guidance: Using Reference Images
Image Guidance (Leonardo's ControlNet implementation) is the feature that makes it most powerful for production work. It allows you to use an existing image to influence the AI's output in specific ways.
Guidance Types
| Guidance Type | What It Controls | Best Use Case |
|---|---|---|
| Pose (OpenPose) | Body position and limb angles | Character in specific action pose |
| Depth | 3D spatial composition | Maintaining scene layout across variations |
| Edge (Canny) | Hard edges and silhouettes | Preserving building/object shapes |
| Style | Visual aesthetic and color palette | Applying art style from a reference image |
| Character Reference | Character appearance/identity | Consistent character across multiple scenes |
Practical Image Guidance Workflow
- Generate or source a pose reference image (can be a simple stick figure or a rendered base mesh)
- Upload it to Leonardo's Image Guidance input
- Set guidance type to "Pose" and influence strength to 0.65-0.85
- Write your character description prompt
- Run generation — your character will adopt the reference pose while the appearance follows your prompt
Character Design Prompts
Character design is where Leonardo AI shines brightest. Here are prompt templates for the most common character types in game production:
Hero/Protagonist
heroic male protagonist, late 20s, short dark hair, determined expression, wearing layered adventure gear with worn leather armor over a simple shirt, sturdy boots, carrying a weathered sword, full body character sheet front view, neutral pose, white background, concept art style, clean lines, highly detailed, digital illustration
Villain/Antagonist
imposing female villain, sharp angular features, silver hair, cold expression, elaborate dark fantasy armor with gold accents and cape, magical energy emanating from her hands, standing pose, full body, dark dramatic background, menacing yet elegant, concept art, masterpiece quality
NPC / Merchant Character
friendly elderly male merchant NPC, warm smile, round spectacles, wearing a colorful vest over a linen shirt, carrying a large pack on his back, approachable and trustworthy, full body character design, fantasy RPG style, warm color palette, game concept art
Creature/Monster Design
original creature design, quadruped predator, obsidian scales with bioluminescent blue markings, four eyes, elongated jaw with multiple rows of teeth, powerful build with blade-like dorsal spines, crouching ready to pounce, full body side view, creature concept sheet, detailed anatomy, fantasy creature art, dark atmospheric background
Environment Concept Art Prompts
Exterior Environment
sweeping fantasy landscape concept art, ancient ruins of a massive stone fortress on a clifftop, overgrown with vines and moss, misty valley below, dramatic late afternoon light breaking through storm clouds, epic scale, painterly style, environment concept art, wide establishing shot, high detail
Interior Environment
underground tavern interior concept art, low stone ceiling with hanging lanterns, long wooden tables, a large fireplace casting warm orange light, barrels along stone walls, several NPCs in the background, cozy and lived-in atmosphere, RPG game environment, detailed architectural features, warm color palette
Sci-Fi Environment
futuristic space station interior corridor, long perspective hallway, metallic walls with glowing blue conduits, holographic displays, subtle ambient lighting, sci-fi game environment concept, clean and corporate aesthetic, cinematic composition, environment art, high detail
Weapon and Item Design Prompts
fantasy sword weapon design, ornate two-handed greatsword, blade engraved with ancient runes that glow faintly gold, wrapped leather grip, crescent-shaped crossguard with inlaid sapphires, full weapon isolated on white background, multiple views (front, side, 45-degree angle), item concept art, game asset design, high detail
magical staff item design, gnarled wood carved from a single piece of driftwood, crystal orb at the top held in twisted wooden fingers, trailing wisps of purple magical energy, isolated on white background, fantasy RPG item art, full view, detailed texture
UI and Icon Art for Games
game UI icon design, health potion, small glass vial filled with glowing red liquid, simple dramatic lighting, isolated on transparent background, 256x256 pixel style, fantasy RPG icon art, clean and readable at small sizes
inventory item icon set concept, six different item icons in a consistent fantasy RPG style: sword, shield, potion, key, scroll, gem, flat lay arrangement, unified warm brown and gold color palette, clean icon art, game UI design
Maintaining Character Consistency Across Scenes
One of the hardest challenges in game art production is keeping the same character looking consistent across multiple illustrations — different poses, scenes, expressions, and lighting conditions. Leonardo offers several approaches:
Style Lock
After generating a character you're happy with, use that image as a Style reference for subsequent generations. Set style influence to 0.5-0.7 to preserve facial features and costume while allowing new poses and settings.
Character Reference Feature
Leonardo's Character Reference (similar to Midjourney's --cref) allows you to upload a character image and use it as a face/identity anchor. The model will attempt to preserve the character's appearance while following your new scene prompt.
Consistent Prompt Anchors
For manual consistency without using reference features, include a consistent "character anchor" phrase at the start of every related prompt:
same female warrior character, dark braided hair, scar over left eye, red cloak, [new scene description]...
Motion Reference and Animation Concepts
character action pose reference, female rogue, mid-air leap, arms outstretched, daggers in both hands, dynamic movement, flowing hair and cloth, strong silhouette readable from 10 feet away, concept art for animation reference, clean background, full body
walk cycle concept sheet, male knight character, four keyframe poses arranged horizontally: contact, down, passing, up phases, consistent character appearance, white background, reference sheet layout, animation concept art
Leonardo's Upscaling and Outpainting
Leonardo's built-in upscaler (Smooth and Creative variants) can take your 512px or 768px outputs to 2048px+ without the quality loss that external upscalers often produce. The Creative upscaler adds new detail during the upscale pass rather than just interpolating existing pixels — useful for adding texture and micro-detail to flat concept art.
Outpainting (Canvas) lets you extend any generated image beyond its original borders — useful for expanding a character portrait into a full environment scene, or turning a landscape thumbnail into a wide-format mural concept.
API Integration for Game Production Pipelines
Leonardo's REST API allows you to build generation requests into custom tools. The basic generation endpoint:
POST https://cloud.leonardo.ai/api/rest/v1/generations
{
"prompt": "your prompt here",
"modelId": "6bef9f1b-29cb-40c7-b9df-32b51c1f67d3",
"width": 1024,
"height": 1024,
"num_images": 4,
"guidance_scale": 7,
"alchemy": true,
"negative_prompt": "blurry, low quality, deformed, ugly"
}
Common production pipeline patterns:
- Batch character variations: Script generates 20 variations of a base character prompt, team reviews and picks directions
- Asset naming and tagging: Script generates, downloads, and names files according to game asset naming conventions
- Automated reference sheets: Given a character description, generate front/side/back/portrait views in sequence
10 Complete Game Art Prompt Examples
-
JRPG Party Member:
cheerful young mage character, teenage girl, bright blue eyes, twin pigtails with star clips, oversized purple robe, holds a glowing magic wand, friendly expression, full body front view, JRPG anime style, clean line art, colorful pastel palette, white background, character design sheet -
Dark Fantasy Boss:
colossal undead dragon boss, skeletal wings stretched wide, cracked bones glowing with necrotic green energy, standing on a pile of ancient ruins, dramatic low angle shot, atmospheric fog, dark fantasy game boss concept art, terrifying and epic scale -
Sci-Fi Soldier:
elite space marine soldier, sleek powered exosuit armor in matte black with orange trim, visor helmet, battle-worn with scorch marks, standing at attention, military bearing, sci-fi game character concept, full body, detailed armor design, studio lighting, white background -
Fantasy Village:
small medieval fantasy village, cobblestone main street, timber-framed buildings with thatched roofs, market stalls, NPCs walking around, warm afternoon light, distant mountain backdrop, establishing shot environment concept, RPG game setting, painterly style -
Dungeon Tileset Concept:
stone dungeon environment tileset concept, three modular tiles shown: straight corridor, T-junction, corner turn, consistent stone brick texture, torch brackets on walls, dungeon RPG game environment art, top-down perspective slightly angled, consistent lighting -
Mobile Game Icon:
mobile game app icon design, fierce dragon head, jewel-toned scales in emerald and gold, breathing fire, dramatic lighting from below, circular icon composition, fantasy mobile game style, bold saturated colors, highly readable at small size -
Stealth Character:
female assassin character, slim build, entirely dark clothing with purple accents, hidden face with glowing purple eyes visible, crouching pose, multiple throwing knives visible on belt, shadow-hugging pose, stealth game character concept, dark atmospheric lighting, full body -
Ancient Artifact Prop:
ancient stone tablet artifact game prop, weathered rectangular stone covered in glowing hieroglyphic symbols, partially broken corner, mysterious golden energy emanating from the cracks, isolated on white background, item concept art, detailed surface texture, archaeological feel -
Vehicle Design:
post-apocalyptic war vehicle design, heavily armored truck mounted with gun turrets, welded scrap metal plates, spiked bumpers, tribal war paint markings, side view technical illustration style, Mad Max aesthetic, detailed mechanical design, vehicle concept art -
UI HUD Concept:
fantasy RPG game HUD interface concept, health bar with ornate dragon scale borders, mana bar with flowing water animation, skill icons in circular slots, inventory quick-access row at bottom, dark semi-transparent panels, dark fantasy UI design, clean readability
Using ImageToPrompt with Leonardo AI
If you have a reference image — a screenshot from a game you admire, a published concept art piece, or a photograph that captures the mood you're going for — ImageToPrompt can extract a structured prompt you can paste directly into Leonardo.
This is particularly powerful for style matching. Upload a screenshot from a game whose visual style you want to reference, and ImageToPrompt will identify:
- The rendering style (painterly, cel-shaded, photorealistic, pixel art)
- Lighting approach (high contrast, soft ambient, dramatic rim lighting)
- Color palette and saturation level
- Level of detail and texture density
Feed that extracted style description into Leonardo alongside your character or environment subject matter. Select the Leonardo model that best matches the style descriptor, and you have a strong foundation for style-matched production art.