Ever had that perfect image in your mind, but struggled to get your AI art tool to actually understand what you're imagining? That’s exactly why we built Tsubaki—our new model that doesn’t just hear your prompts, but truly gets them.

Let's dive into some real examples that demonstrate how our model pushes the boundaries of AI-generated imagery, turning complex creative visions into stunning visual reality.

Tsubaki's Advanced Generation Capabilities

What really sets this model apart is how it transforms your prompts into fully-realized worlds, not just individual images, but cohesive visual stories with remarkable consistency and depth.

Character Consistency Engine

Building on our multi-character techniques, Tsubaki maintains flawless continuity when you need:

• The same character across multiple scenes
• Consistent outfits and accessories
• Stable facial features in different expressions

Why Consistency Matters? This is extremely helpful for comic creations, as creating a comic requires your characters to look identical across hundreds of panels - something incredibly difficult with most AI models.

Narrative Intelligence

Tsubaki demonstrates exceptional narrative intelligence through its ability to process and execute complex, detailed instructions while maintaining visual coherence across multiple elements.

1. Character-Environment Synergy

Tsubaki excels at creating meaningful interactions between characters and their surroundings, going beyond simple placement to achieve true environmental storytelling.

2. Composition Innovation

One of Tsubaki's most impressive capabilities is its mastery of unconventional compositions and camera angles that break away from standard AI-generated imagery patterns.

3. Enhanced Lighting and Shadow Effects

Beyond compositional mastery, Tsubaki demonstrates sophisticated lighting simulation through advanced algorithms for dynamic light diffusion, ambient occlusion, and natural shadow blending.

4. Real-World Knowledge

Tsubaki also demonstrates understanding of cultural references, historical contexts, and real-world physics that goes beyond pattern matching.

Mastering Tsubaki: Your Complete Prompt Engineering Guide

Now that you've seen what Tsubaki can accomplish, you're probably wondering how to harness these capabilities for your own creative projects. The secret to unlocking Tsubaki's full potential lies in understanding how to craft effective prompts that leverage its advanced features.

Prompt Helper – Your Secret Weapon

First up: Meet Prompt Helper. Think of it as your behind-the-scenes writing partner, quietly refining your prompts to help Tsubaki deliver even better results.

It works with any input style - whether you prefer simple tags, short natural language, mixed styles, or even prompts in your native language, Prompt Helper bridges the gap, making sure the model understands exactly what you’re going for.

For example, let's start with a simple tag like this: 1girl, black hair, short hair, school uniform, smiling

With Prompt Helper, it transforms into something like:

• Short natural language:

• Mixed style(tag+natural language)

• Prompts in non-English language: 桜の下で和服を着た女性

See the difference? The enhanced version gives Tsubaki way more context—pose details, clothing textures, even the energy of the scene. And when we compare outputs, the Prompt Helper version consistently delivers more dynamic, polished results.

Want to see magic in action? After generating, check the Generation Task section—you’ll see exactly how Prompt Helper enhanced your original input.

Natural Language Prompting with Tsubaki

Imagine you're describing this beautiful scene to a friend: 'A young girl in a white sundress walking through a sunflower field, with that golden afternoon light making everything look warm and nostalgic, like a watercolor painting come to life.'

Now, with older AI systems, you'd have to break that down into this robotic list: '1girl, full_body, walking_motion, sunflower_field, surrounded_by_flowers, sunlight, golden_hour, summer, nostalgic_style, watercolor_effect' and even then you might not get that special feeling you wanted. But Tsubaki? It gets it. It understands that she's moving through the flowers, not just standing in front of them.

A practical approach to writing effective prompts:

First, we'll start by establishing our main subject and what she's doing - this forms the foundation of our scene. Here, we have 'A blushing schoolgirl nervously adjusts her glasses with one hand while clutching a love letter in the other.'

Next, we'll build on this foundation by adding descriptive details about her appearance. In this case, we're including 'her long lavender hair fluttering slightly in the spring breeze.' These visual details help Tsubaki understand exactly how our character looks and moves in the scene.

Now we'll establish the environment and lighting conditions to set the stage. We're adding 'Cherry blossom petals dance around her as she stands frozen outside the classroom door.' This places our character in a specific location with clear seasonal elements that will influence the lighting and atmosphere.

For important interactions, we want to highlight how the character engages with her surroundings. In this example, we're emphasizing that she's 'mustering the courage to go inside,' which explains her relationship to the classroom door and suggests an upcoming action.

Finally, we've naturally incorporated the overall mood through word choices like 'blushing,' 'nervously,' and 'mustering the courage.' These emotional cues help Tsubaki understand the tender, anxious atmosphere we're trying to create.

Putting it all together, we get a complete scene description that flows naturally while covering all the essential elements: 'A blushing schoolgirl nervously adjusts her glasses with one hand while clutching a love letter in the other, her long lavender hair fluttering slightly in the spring breeze. Cherry blossom petals dance around her as she stands outside the classroom door, mustering the courage to go inside.'

Multi-Character Scenes Made Easy

Now let's talk about bringing multiple characters to life in a single scene - something that used to be incredibly tricky with AI art tools. What makes Tsubaki special is how it understands these complex interactions when you describe them naturally. Take this scene of two teens by a vintage car.

The magic starts with clear positioning. Notice how we immediately establish where everyone is: 'A red-haired girl sits on the hood... a black-haired boy stands beside the car.' This gives Tsubaki the foundation.

Next comes defining each character visually. For the girl, we don't just say 'school uniform' - we specify 'black serafuku with sailor collar, pleated miniskirt, knee-high black socks'. For the boy, it's not just 'jacket and pants' but 'brown jacket over black shirt, black pants, glasses, grey scarf loosely draped'. These specific details act like visual fingerprints - Tsubaki uses them to keep characters distinct.

But what really brings the scene alive are the interactions and small moments. She's not just holding a drink - she's 'smiling gently as she looks at the boy'. He's not just standing - he's 'leaning casually against the car, hands in jacket pockets'. These subtle behaviors tell Tsubaki how they relate to each other and their environment.

Finally, we tie it together with atmosphere and mood. Those 'autumn leaves scattering around them' and 'long shadows in the afternoon light' create emotional texture. The mention of 'warm tones against cool blue' guides the color palette, while 'nostalgic autumn mood' sets the emotional tone.

Next, let's tackle group shots which require every character to maintain crystal-clear individuality. The key is still, giving each person unique visual signatures across multiple categories. Here's an example of how to improve a school scene:

We begin with this prompt:

The initial description missed opportunities to differentiate characters through specific hair/eye color combinations, detailed outfit variations, distinctive accessories or unique posing cues. As we only described their roles (e.g., "class president," "athletic team captain"), Tsubaki might default to similar outfits or features. Instead, we need to define unique visual traits for each character—hair, eyes, clothing, accessories, and poses.

Let's revise this prompt with more details.

This version had three issues:

1. The volleyball captain's pose was physically awkward (impossible knee position, also the looker's position seemed weird in the scene)
2. The height differences made them seem like different age groups
3. The spatial relationships weren't quite natural

To fix this, let's revise the prompt again:

First, we established consistent proportions by specifying:

"All three characters are of the same age and almost the same height."

This simple line ensures Tsubaki won't accidentally create unrealistic size variations between classmates.

Then, we refined the captain's pose from an exaggerated stance to something more natural:

Before: "She's got one knee bent with her foot propped against the lockers"

After: "She's leaning against the lockers"

This maintains her athletic personality while being physically plausible.

Final outcome:

Refinement techniques

1. Strengthen the spatial relationships

Original Prompt: "Two boys feeding a cat outside"

Improved Prompt:

Low-angle shot dominated in the foreground by a curious black cat, its tail curling around fallen autumn leaves as two teenage boys kneel on the sidewalk behind it. The orange-haired boy leans forward with an outstretched hand offering a treat, his unzipped black jacket revealing a rumpled white shirt underneath. Beside him, the black-haired boy watches through narrowed eyes, one hand resting on his knee while clutching a half-open food wrapper. Dappled sunlight filters through overhead trees, catching the dust motes around their playful interaction near a yield sign.

2. Enhance the distinguishing features

Original Prompt: "Twin schoolgirls in uniforms"

Improved Prompt:

Two teenage girls stand side by side in matching school uniforms, looking directly at the viewer with contrasting expressions. The girl on the left has short blue hair with colored inner streaks styled in a bob cut, featuring aegyo sal under her bright grey eyes. She wears a black serafuku with a white collared shirt, aqua necktie, and black pleated skirt, layered under a green jacket with sleeves extending past her wrists. Her companion has medium-length grey hair in a wolf cut with multicolored streaks, frowning slightly with her mouth closed while wearing an identical uniform with an aqua bow instead of a necktie. Both girls' black sweaters peek from beneath their jackets, and their sleeves drape past their fingers.

3. Add interaction cues

Original Prompt: "Two girls holding hands"

Improved Prompt:

Two young women stand face-to-face in an intimate pose, their upper bodies visible against a clean white background. The first girl has long blonde hair styled in a low side ponytail, adorned with a green ribbon and delicate jingle bells. She wears black-framed glasses and a green camisole that reveals her bare shoulders and collarbone. Her green eyes gaze softly at the other girl, one hand resting gently on her companion's shoulder. The second girl has short purple hair styled in twintails, with a grey camisole and matching hair ribbon. A small bell ornament decorates her hair between her eyes. She raises one hand up while interlocking fingers with the blonde girl's other hand, creating a tender yuri atmosphere. Both wear subtle hair ornaments and maintain closed mouths with serene expressions. The simple composition focuses on their physical connection through held hands and shoulder contact, emphasizing their quiet emotional bond through body language rather than overt expressions.

• Use clear positional words (opposite, behind, between)
• Give each character distinct visual signatures
• Include action verbs that differentiate them
• Establish relationships through their interactions
• When troubleshooting, first clarify who is where doing what

With practice, you'll be able to create complex group shots where every character maintains their unique identity while contributing to a cohesive scene. The more specific you are about how characters relate to each other spatially and emotionally, the better Tsubaki can bring your vision to life.

Final thoughts

That's a wrap on our deep dive into Tsubaki's capabilities! Whether you're creating comics, crafting detailed fantasy worlds, or exploring innovative visual storytelling, Tsubaki gives you the tools to bring your most ambitious creative visions to life.

To celebrate Tsubaki's official launch, we're hosting an exciting contest on our official Twitter account! From July 10th to July 24th, 2025, we're inviting all creators to showcase what they can accomplish with Tsubaki. We've prepared amazing prizes including Amazon gift cards and 200,000 credits, plus every participant will receive a special participation reward! Don't miss out!

We can't wait to see the incredible art you'll create with Tsubaki. Happy creating!

Updated on: 10/07/2025