Pixel art is easy to recognize, but getting an AI image generator to reproduce the style consistently can be surprisingly difficult.
Writing only pixel art in a prompt may give you something that looks pixelated, but the result can still resemble a normal anime illustration with a pixel filter applied on top.
A better approach is to describe the visual characteristics you actually want: stepped outlines, grouped colors, reduced shading, simple shapes, and visible pixel blocks.
In this tutorial, I’ll use Tsubaki.3 on PixAI to create a full-body pixel-style character, compare a basic prompt against a more detailed one, and test a small prompt adjustment to see whether it produces a stronger retro-game look.
The goal is a pixel-style illustration, not a production-ready game sprite. If you need a fixed sprite size, transparent background, animation sheet, or exact pixel grid, the generated image may still need manual editing afterward.
1. Select Tsubaki.3
Start by opening PixAI's image generation interface and selecting Tsubaki.3.
For this test, I kept the generation setup simple because I wanted to focus mainly on how the prompt affects the style.
My setup:
Model: Tsubaki.3
Image size/aspect ratio: 3:5
Other settings changed: None
Use the settings from your actual generation rather than assuming values, since the interface and available controls can change.
2. Build a Reusable Pixel-Art Prompt
Instead of writing every prompt from scratch, I started with a reusable structure.
Pixel art illustration of [character type], full body,
[hairstyle and hair color], [clothing and accessories],
[pose or action].
Centered on a plain [background color] background.
Retro 2D RPG character aesthetic,
clearly visible square pixel blocks,
crisp stair-stepped outlines,
flat color clusters,
limited color palette,
minimal shading,
simple readable silhouette.
No text.
The first half describes the subject.
The second half controls the visual treatment.
For my first test, I created a fantasy courier:
Pixel art illustration of a young female fantasy courier,
full body, short dark-red bob hair, cream tunic,
brown shorts, ankle boots, and a small green shoulder bag,
standing while holding a folded letter.
Centered on a plain pale beige background.
Retro 2D RPG character aesthetic,
clearly visible square pixel blocks,
crisp stair-stepped outlines,
flat color clusters,
limited color palette,
minimal shading,
simple readable silhouette.
No text.
I intentionally kept the character simple: one person, a plain background, limited accessories, and a clear full-body pose. This makes it easier to judge whether the pixel treatment is actually working.
3. Generate the First Character
After generating the image, I checked two things:
- whether the character matched the description;
- whether the image actually looked pixel-styled.
The result followed the character description closely. The dark-red hair, cream tunic, brown shorts, boots, green shoulder bag, and folded letter were all visible.
The pixel treatment was also noticeable around the hair, clothing, boots, and accessories.
Instead of completely smooth curves, many edges had a stepped appearance. The colors were also grouped into clearer blocks, while the shading was simpler than what I would expect from a highly rendered anime illustration.
However, the image still contained a fair amount of detail in the face, hair, hands, and clothing folds.
That is why I would describe it as pixel-style artwork rather than strict traditional pixel art.
The prompt can influence the visual appearance, but it does not guarantee that the image follows a fixed low-resolution pixel grid or uses an exact palette.
4. Does a Longer Pixel-Art Prompt Actually Help?
To test the style wording, I generated the same character again using a much simpler prompt:
Pixel art illustration of a young female fantasy courier,
full body, short dark-red bob hair, cream tunic,
brown shorts, ankle boots, and a small green shoulder bag,
standing while holding a folded letter.
Centered on a plain pale beige background.
No text.
The character description and background stayed the same.
The main difference was that I removed:
Retro 2D RPG character aesthetic,
clearly visible square pixel blocks,
crisp stair-stepped outlines,
flat color clusters,
limited color palette,
minimal shading,
simple readable silhouette.
In this test, using the phrase "pixel art illustration" was already enough for Tsubaki.3 to produce something recognizably pixel-styled.
However, the simpler version looked slightly closer to a polished anime illustration with pixel-like rendering.
The longer prompt produced a stronger retro RPG feel. The difference was not dramatic, but the shapes, outlines, and grouped colors looked more deliberate.
What Do the Extra Prompt Terms Actually Do?
Each phrase targets a different part of the rendering.
clearly visible square pixel blocks
Encourages a more obviously block-based appearance.
crisp stair-stepped outlines
Pushes curved edges toward visible pixel-like steps rather than smooth lines.
flat color clusters
Encourages larger areas of grouped color instead of continuous blending.
limited color palette
Reduces excessive color variation.
minimal shading
Discourages complex gradients and highly rendered lighting.
simple readable silhouette
Keeps the overall character shape easy to recognize.
These instructions do not technically enforce pixel-art rules, but they give the model more information about the visual direction you want.
5. Push the Result Further Toward a Retro Look
The first result already worked reasonably well, so I made one small adjustment rather than rewriting the entire prompt.
I added:
Larger visible pixel blocks,
simpler shading,
fewer fine details.
The full prompt became:
Pixel art illustration of a young female fantasy courier,
full body, short dark-red bob hair, cream tunic,
brown shorts, ankle boots, and a small green shoulder bag,
standing while holding a folded letter.
Centered on a plain pale beige background.
Retro 2D RPG character aesthetic,
clearly visible square pixel blocks,
crisp stair-stepped outlines,
flat color clusters,
limited color palette,
minimal shading,
simple readable silhouette,
larger visible pixel blocks,
simpler shading,
fewer fine details.
No text.
The result changed, although not dramatically.
The image became slightly more blocky, particularly around the hair and clothing edges, while some smaller details were reduced.
That made it feel closer to a retro game-style illustration.
The important point is that prompt adjustments like this influence the result. They do not guarantee that every generation will suddenly become a technically accurate sprite.
6. A Simple Way to Troubleshoot Pixel-Style Outputs
If your result looks too smooth, try adding:
Larger visible pixel blocks,
simpler shading,
fewer fine details.
If the character becomes too complicated, remove some clothing details or accessories.
If the background distracts from the character, add:
No additional background objects.
The best approach is to change one variable at a time.
If you modify the character, background, lighting, palette, and pixel instructions simultaneously, it becomes difficult to tell which change actually improved or worsened the output.
7. Reuse the Style for Another Character
Once I had a prompt structure that worked, I tested it with a different character.
This time, I created an apprentice alchemist:
Pixel art illustration of a young male apprentice alchemist,
full body, messy dark blue hair, light tunic,
short brown cape, dark trousers, boots,
and a small belt pouch, holding a small bottle.
Centered on a plain pale gray-blue background.
Retro 2D RPG character aesthetic,
clearly visible square pixel blocks,
crisp stair-stepped outlines,
flat color clusters,
limited color palette,
minimal shading,
simple readable silhouette.
No text.
The character design changed, but the style instructions remained almost identical.
The generated image still maintained the block-like outlines and grouped colors, even though the hair, clothing, accessories, and pose were different.
That is the main advantage of using a reusable prompt template.
Once the style portion works reasonably well, you can leave it alone and modify only the subject-specific variables.
8. Export the Image
After generating the final image, record the actual export properties.
File format: PNG
Dimensions: 1104 x 1824
Use the values from the downloaded file rather than guessing them.
The generated artwork can work well for things like:
- character concepts;
- profile images;
- pixel-style illustrations;
- visual references;
- social content.
But don't assume that an AI-generated pixel-style image is automatically suitable as a game asset.
A production sprite may require additional work such as:
Fixed sprite dimensions
Transparent background cleanup
Palette correction
Manual pixel editing
Animation frames
Sprite-sheet preparation
The generated result is better treated as a starting point unless you have verified the technical requirements of your project.
Final Thoughts
The useful part of this experiment wasn't finding one perfect pixel-art prompt.
It was learning which instructions changed the rendering.
My workflow was:
Select Tsubaki.3
↓
Fill in the character template
↓
Generate
↓
Inspect the pixel treatment
↓
Change one prompt detail
↓
Generate again
↓
Reuse the working style
Tsubaki.3 already responded reasonably well to the phrase "pixel art illustration", but adding more specific instructions gave me better control over the outlines, color grouping, shading, and amount of detail.
And when the first result still looked slightly too polished, adding:
Larger visible pixel blocks,
simpler shading,
fewer fine details.
gave me a simple way to push the image further toward the retro look I wanted.
Once you find a style description that works, you can keep that part of the prompt mostly unchanged and experiment with different characters, clothing, poses, and backgrounds instead.
Top comments (0)