Table of Contents
- Why Multiple Character Scenes Are So Hard for AI
- The Solution: Multiple Character Consistency V2
- Getting Started with Multi-Character V2
- Step 1: Create Perfect Reference Images First
- Step 2: Use the 3-Step Prompting Method
- Part 1: Define Your Character Tags First
- Part 2: Set Positions Before Actions
- Part 3: Then Add Actions and Interactions
- Step 3: Set the Background
- Complete Example Prompts
- Common Mistakes to Avoid
- Mistake 1: Cramming everything into one sentence
- Mistake 2: Being vague about positions
- Mistake 3: Overcomplicating the prompt
- Mistake 4: Using low-quality reference images
- Mistake 5: Mixing character descriptions with scene descriptions
- Why This Method Works
- Advanced Tips for Professional Results
- Tip 1: Create a scene library
- Tip 2: Match your aspect ratios
- Tip 3: Build complexity gradually
- Tip 4: Maintain lighting consistency
- Tip 5: Regenerate when needed
- The Complete Workflow Summary
- Start Creating Your AI Cartoon Stories Today
- Frequently Asked Questions
Do not index
Do not index
Creating scenes with multiple cartoon characters that maintain perfect consistency across different images has always been one of the biggest challenges in AI-generated art. Most AI tools struggle with this fundamental problem, but there's now a breakthrough solution that actually works.
If you've ever tried creating AI illustrations with two characters interacting, you know the frustration. You get one character looking perfect. Then you add a second character to the scene and everything falls apart.
Hair colors change. Outfits morph. Facial features drift. By the third scene, your characters look like completely different people.
You're not alone. This is the number one complaint we hear from children's book creators: "I can get one character to look great, but the moment I try to add a second character to the scene, everything falls apart."
After months of testing different approaches and analyzing thousands of failed attempts, we cracked the code. And today, we're sharing exactly how to master multiple character consistency for your AI storybook illustrations.
Why Multiple Character Scenes Are So Hard for AI
Before we dive into the solution, it helps to understand why this problem exists in the first place.
When you generate an image with a single character, the AI can focus all its attention on maintaining that character's appearance. Add a second character, and suddenly the AI has to juggle two distinct visual identities simultaneously while also composing a coherent scene.
Most AI systems weren't built for this. They treat each generation as independent, with no memory of what your characters looked like in previous images. And when you try to describe two characters in one prompt, the AI often gets confused about which features belong to which character.
The result? Characters that swap features, merge appearances, or simply look different from scene to scene. We cover this problem in depth in our guide on how to keep AI characters consistent, but multi-character scenes add an extra layer of complexity.
The Solution: Multiple Character Consistency V2
Neolemon has developed Multiple Character Consistency V2, which combines character reference technology with a simple 3-step prompting method that delivers reliable results. This system allows you to create new scenes with full character consistency every time.
The breakthrough came when we combined two approaches: using high-quality reference images AND structuring prompts in a very specific way.
Neither works well alone. Reference images without proper prompting still produce inconsistent results. And even the best prompts can't overcome the limitations of low-quality reference images.
But together? They produce multiple character scenes where both characters stay remarkably consistent across dozens of illustrations.
Here's the complete workflow.
Getting Started with Multi-Character V2
First, you'll need to sign up for Neolemon, where you get 20 credits for free to try the service. Once you're in the platform, click on Multi-Character in the home dashboard to access the multiple character feature.
Step 1: Create Perfect Reference Images First
Everything starts with your character reference images. These are the foundation that every scene builds upon.
Upload your characters to the platform. For best results, these should be full body front view images of your cartoon characters. The system will automatically label them as @character1 and @character2, though you can customize these labels by clicking on them (for example, @amy and @leo).
For each character, you need a high-quality image that shows:
Full body view: The AI needs to see the complete character, from head to toe. This includes their proportions, posture, and how their outfit fits their body.
Front-facing pose: A clear, straight-on view works best. Dynamic angles can be added later, but your reference should show the character in a simple, readable pose.
Clean background: A solid color or simple gradient keeps the focus on the character. Busy backgrounds can confuse the AI about what elements belong to the character versus the environment.
Consistent style: Both characters should match in artistic style. Mixing a watercolor character with a Pixar-style character will create visual chaos in your scenes.
Don't have reference images yet? You can create them using Character Turbo, which is specifically designed to generate clean, front-facing character references that work perfectly for multi-character scenes. Watch our step-by-step Character Turbo tutorial to see exactly how it works.
Pro tip: Spend extra time getting your reference images right. It's much easier to regenerate a reference image than to fix consistency problems across 20+ story illustrations.
.png?table=block&id=2d8ce913-3b4a-80d9-b2b6-d7bad8879b13&cache=v2)
Step 2: Use the 3-Step Prompting Method
The key to success lies in following this exact sequence when creating your prompts. We discovered that AI needs structure, not complexity. Most people try to describe everything in one long, rambling sentence. The AI gets confused, and the results suffer.
Part 1: Define Your Character Tags First
Always start your prompt by telling the AI model who's who. This establishes clear identities before anything else happens.
Examples:
- "@character1 is a girl and @character2 is a boy"
- "@character1 is a dog and @character2 is a child"
- "@amy is a girl and @leo is a boy"
This seems simple, but it's crucial. Without clear tags, the AI might assign the wrong actions or positions to the wrong character.
Part 2: Set Positions Before Actions
Before describing what your characters are doing, tell the AI exactly where each character should be in the scene. Spatial relationships need to be established before adding complexity.
Examples:
- "@character1 is standing in the front, @character2 is behind and to the left of @character1"
- "@character1 is sitting next to @character2"
- "@character1 is standing to the left and @character2 is standing to the right"
This step prevents the common problem of characters appearing in unexpected positions or overlapping awkwardly.
Part 3: Then Add Actions and Interactions
Only after establishing identity and position do you layer in what the characters are doing. This is where you can get creative with expressions, gestures, and interactions.
Examples:
- "@character1 is looking at @character2 with a surprised expression"
- "@character1 is giving a present to @character2"
- "@character1 and @character2 are talking and smiling"
- "@character2 is jumping and giving a high five to @character1"
Step 3: Set the Background
Describe your background in simple, vivid terms. The AI works best with clear instructions. Use the separate Background field in the interface for best results.
Examples:
- "a magical forest"
- "in a beautiful park"
- "a cozy living room with warm lighting"
- "a sunny beach at sunset"
Complete Example Prompts
Here's what complete prompts look like in practice:
Example 1: High-five in a magical forest
Prompt: @character1 is a girl and @character2 is a boy. @character1 is standing to the left and @character2 is standing to the right. @character2 is jumping and giving a high five to @character1
Background: a magical forest
Example 2: Sharing ice cream in the park
Prompt: @character1 is a girl and @character2 is a boy. @character1 and @character2 are sitting together. They are laughing and sharing ice cream.
Background: in a beautiful park
Example 3: Dynamic adventure scene
Prompt: @character1 is the main character and @character2 is the sidekick. @character1 is in the foreground running forward, @character2 is following close behind. @character1 is reaching out toward something off-screen with an excited expression, @character2 looks worried but determined.
Background: a mysterious cave entrance
Example 4: Emotional storytelling moment
Prompt: @character1 is the child and @character2 is the grandmother. @character1 is sitting in @character2's lap. @character1 is looking up at @character2 with wonder while @character2 smiles warmly and reads from a storybook.
Background: a cozy bedroom at night with soft lamp light
Notice how each example follows the same pattern: tags, positions, then actions. This consistency in your prompting leads to consistency in your results.
For a deeper dive into prompting techniques, check out our complete masterclass on AI cartoon storybook illustrations, which covers everything from character DNA to scene composition.
Common Mistakes to Avoid
Even with the right method, a few common mistakes can derail your results:
Mistake 1: Cramming everything into one sentence
Wrong: "@character1 who is a boy standing on the left is waving at @character2 who is a girl on the right while they both smile in a park setting with trees."
Right: Break it into the 3-step structure. Tags first, positions second, actions third.
Mistake 2: Being vague about positions
Wrong: "@character1 and @character2 are in the scene together."
Right: "@character1 is standing on the left side. @character2 is sitting to the right of @character1."
Mistake 3: Overcomplicating the prompt
Wrong: A 200-word description trying to capture every detail of a complex scene.
Right: Keep it simple. Focus on the key elements that matter for your story. The AI handles the rest.
Mistake 4: Using low-quality reference images
If your reference images are blurry, cropped awkwardly, or show inconsistent styling, no prompting technique will save you. Invest time upfront in creating clean, high-quality references.
Mistake 5: Mixing character descriptions with scene descriptions
❌Wrong: "The boy with brown hair in a blue shirt is standing in a sunny meadow with flowers while the girl with blonde hair..."
✅Right: Let your reference images carry the character descriptions. Your prompt focuses on positions and actions. The background goes in the separate Background field.
Why This Method Works
The systematic approach removes guesswork from multi-character scene creation. By following the exact sequence of define tags → set positions → add actions, you give the AI the information it needs in the right order, resulting in consistent, high-quality character interactions across all your generated scenes.
Beta users who tested this workflow are telling us it's "one of the best multiple character consistency tools they've used in a long time" and that "it's very clever in understanding natural prompts, and the images are super high quality and consistent."
Advanced Tips for Professional Results
Once you've mastered the basics, these advanced techniques can take your multi-character scenes to the next level:
Tip 1: Create a scene library
After generating a particularly good scene, save it. You can use successful outputs as additional reference material for future generations in similar settings.
Tip 2: Match your aspect ratios
If your children's book uses square pages, generate square images using the Aspect Ratio selector. Landscape scenes stretched into square formats (or vice versa) can distort your characters' proportions.
Tip 3: Build complexity gradually
For complex scenes with specific interactions, sometimes it helps to generate a simpler version first, then iterate toward the final vision. Going from "standing together" to "playing catch" to "playing catch while laughing in the rain" in stages often produces better results than jumping straight to the complex version.
Tip 4: Maintain lighting consistency
Your book will look more professional if lighting stays consistent across scenes. If scene one has warm afternoon lighting, carrying that through to subsequent outdoor scenes creates visual coherence beyond just character consistency.
Tip 5: Regenerate when needed
Not every generation will be perfect. That's normal. The 3-step prompting method dramatically improves your hit rate, but don't hesitate to regenerate if something looks off. It's always faster than trying to fix issues in post-production.
If you're working on a children's book specifically, you might also find our guide on how to create children's books with AI using Neolemon helpful for the complete end-to-end workflow from character creation to KDP publishing.
The Complete Workflow Summary
Let's recap the entire process:
- Sign up and access Multi-Character: Get 20 free credits and click Multi-Character in the dashboard
- Upload reference images for each character (full body, front view, clean background, matching style)
- Customize character tags by clicking on @character1 and @character2 labels (optional but helpful)
- Write your prompt using the 3-step structure:
- Step 1: Define tags ("@character1 is a girl and @character2 is a boy")
- Step 2: Set positions ("@character1 is on the left, @character2 is...")
- Step 3: Add actions ("@character1 is doing X, @character2 is doing Y")
- Set your background in the separate Background field
- Choose aspect ratio (Square 1:1 for most children's books)
- Generate and iterate until you get the scene you need
- Repeat for each scene in your storybook
Start Creating Your AI Cartoon Stories Today
Ready to bring your characters to life with perfect consistency? Join thousands of creators who are already using Neolemon to create beautiful AI cartoon stories for children's books, illustrations, and visual narratives.
The combination of high-quality reference images and structured 3-step prompting produces results that simply weren't possible with other methods. Both characters maintain their distinct appearances across scene after scene.
Get started now with 20 free credits and experience the difference that true character consistency makes in your storytelling. Whether you're a children's book author, illustrator, or content creator, Neolemon gives you the tools to create professional-quality cartoon scenes that maintain perfect character consistency across every image.
Sign up for Neolemon and start creating your first multi-character scene in minutes. No credit card required.
Frequently Asked Questions
How many characters can appear in one scene?
Currently, Multi-Character V2 is optimized for two characters in a single scene. We're actively working on supporting more than two characters in future updates.
How many credits does each generation cost?
Each Multi-Character generation uses credits from your account. You get 20 free credits when you sign up, which is enough to test the workflow and create several scenes.
Do I need to create new reference images for each project?
No. Once you have high-quality reference images for your characters, you can use them across unlimited scenes and projects. That's the beauty of the reference-based approach.
What if my characters need to change outfits?
You have two options: create additional reference images showing the outfit change, or describe the outfit modification in your prompt. For significant outfit changes (like switching from casual clothes to a costume), new reference images typically produce better results. Learn more about outfit changes in our Expression and Outfit Editor guide.
Can I use characters I created in other AI tools?
Yes. As long as you have a clear, full-body, front-facing image of your character, you can use it as a reference. However, characters created specifically in Neolemon's Character Turbo tend to produce the most consistent results since they're optimized for our multi-character system.
What aspect ratio works best?
It depends on your book format. Square (1:1) works well for many children's books. Portrait (3:4) is great for standard book pages. Landscape (4:3) suits spread illustrations. Use the Aspect Ratio selector to match your final book dimensions.
Can I customize the character tag names?
Yes! Click on @character1 or @character2 in the interface to rename them to something more memorable like @amy and @leo. This makes writing prompts more intuitive.
Where can I learn more?
Join our free Neolemon community for tutorial videos, live workshops, office hours, and regular challenges where you can earn free credits for sharing your creations.
Written by
