> For the complete documentation index, see [llms.txt](https://docs.seaart.ai/guide-1/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.seaart.ai/guide-1/seaart-ai-basic-function/text-to-image.md).

# Text to Image

Learn what is text to image and how to use AI text-to-image generator with easy, step-by-step instructions.

> **Have you ever encountered similar problems while using SeaArt: the control effect is not ideal, or the drawing results do not reflect the added prompts, and so on? In this article, we will comprehensively introduce how to master the operation of text-to-image and grasp the strategy of writing efficient prompt words.**

## What is Text to Image?

In SeaArt AI, there are three modes of drawing: <mark style="color:red;">Text to Image</mark>, [<mark style="color:red;">Image to Image</mark>](/guide-1/seaart-ai-basic-function/image-to-image.md)**,** and [<mark style="color:red;">ControINet</mark>](/guide-1/seaart-ai-basic-function/controlnet.md)

The basic steps for drawing are: <mark style="background-color:$warning;">select a model→ enter prompts→ set parameters→ generate.</mark>

The model determines the style, prompts define the content of the image, and parameters refine the preset characteristics of the image.

### Create your first image

1. Open [SeaArt AI](https://www.seaart.ai/) and choose a text-to-image model that fits your idea, such as a photo, anime illustration, or product image.
2. Enter a prompt describing what you want to see. Be clear about the subject and the scene. Add details like lighting or framing if they matter.
3. Choose an available aspect ratio or image size, then check the displayed generation cost.
4. Generate the image and review the result. If something looks off, change that part of the prompt and try again.

{% prompt description="Example: Cat by a window" %}

```markdown
A small orange cat sits on a wooden windowsill beside a white ceramic mug. Soft morning light comes through the window from the left. The cat is in focus, with a garden softly blurred in the background. Natural photography.
```

{% endprompt %}

### How to write a clear prompt

Think of your prompt as a short visual brief. Say what should appear, where it is, and what you want the image to look like. One or two specific sentences are often enough to get started.

For example, instead of writing “beautiful coat, best quality, amazing details,” describe the visible details you care about:

{% prompt description="Example: Anime Portrait" %}

```markdown
An adult woman sits in front of a circular wooden lattice window, wearing a solid orange coat over a white top and gray trousers. The entire coat is orange, including the sleeves. Soft daylight, centered portrait, detailed anime illustration.
```

{% endprompt %}

A few practical tips:

* **Be specific.** “Soft window light from the left” gives clearer direction than “nice lighting.”
* **Keep details consistent.** If the coat should be orange, don't also describe it as yellow.
* **Don't add details just to make the prompt longer.** Include what matters to the image and leave out repeated or conflicting instructions.
* **Choose one clear style.** For example, use “studio product photograph” or “watercolor illustration” to set the overall look.

## Prompt examples

Use these as starting points and change the subject or details to fit your idea.

### Character illustration

{% prompt description="Prompt: Character and setting" %}

```markdown
An adult traveler in a blue coat walks through a narrow fantasy city street. Show the traveler from head to toe, with tall buildings on either side. Soft morning light falls across the cobblestones. Anime illustration with clean linework and muted colors.
```

{% endprompt %}

### Product photo

{% prompt description="Prompt: Product and lighting" %}

```markdown
A matte green ceramic bottle stands on a pale stone surface against a plain cream background. Soft light from the upper left creates a gentle shadow to the right. Front three-quarter view, studio product photography.
```

{% endprompt %}

### Landscape

{% prompt description="Prompt: Scene and atmosphere" %}

```markdown
A mountain lake at dawn, with a small wooden jetty entering from the lower left. Mist hangs above the water and distant peaks catch the first warm light. Wide landscape photograph with muted colors.
```

{% endprompt %}

## How to improve the result

When an image isn't quite right, identify the main issue and change that part of the prompt. You don't need to rewrite everything.

* **The subject is cropped:** ask for a full-body view and leave space around the subject. Check that the aspect ratio fits your image.
* **The background is too busy:** describe a simpler setting, such as “a plain cream wall with no other objects.”
* **The colors are wrong:** state the intended color clearly and remove any conflicting color descriptions.
* **The style feels off:** try a more specific style description or choose a model better suited to the result you want.
* **You need to keep a specific face, product, or layout:** use a suitable reference-image or image-editing workflow. A text prompt alone may not preserve exact details.

*<mark style="color:$primary;">**Tip:**</mark> <mark style="color:$primary;"></mark><mark style="color:$primary;">Change one thing at a time when possible. It makes it easier to see which instruction helped.</mark>*

## Prompt weights and negative prompts

Natural-language prompts are a good starting point, but some models also support prompt weights such as `(orange coat:1.2)`, or use parentheses and square brackets to adjust emphasis. These options are model-dependent, so check the selected model's guidance before using them.

Negative prompts can also help exclude unwanted elements, such as blurry details or objects you don't want in the image. Their availability and effect depend on the model and its settings.

For examples and a more detailed explanation of prompt weighting, syntax, and negative prompts, see the [Complete Prompting Guide](https://docs.seaart.ai/guide-1/parameters/complete-prompting-guide).

### Older syntax and natural-language prompts

For example, an older prompt might use <mark style="color:$danger;">`yellow coat, (((orange coat)))`</mark> to emphasize orange. A natural-language alternative is: “The entire coat is solid orange, including the sleeves.”

<figure><img src="https://2219884424-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FigAtVLBrlaI8jVruJfC8%2Fuploads%2FTiYYtrJGVFLrnhatH9To%2F%E5%A4%96%E5%A5%97%E6%9B%B4%E6%96%B0.webp?alt=media&amp;token=07b6759a-5f0a-4de7-90d9-af3559069c8e" alt=""><figcaption></figcaption></figure>

Another older prompt might use <mark style="color:$danger;">`((yellow coat)), [[orange coat]]`</mark> to emphasize yellow and reduce orange. A natural-language alternative is: “The entire coat is mustard yellow, including the sleeves.”

<figure><img src="https://2219884424-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FigAtVLBrlaI8jVruJfC8%2Fuploads%2F2HzCsRW2Hk7nBfUF5MN2%2F%E6%96%87%E6%A1%A3%E6%9B%B4%E6%96%B02.webp?alt=media&amp;token=372b802c-44b9-4584-88ee-dd9f93890eda" alt=""><figcaption></figcaption></figure>

Both approaches may be useful in the right context. The symbol-based version depends on model support, while the sentence explains the intended result directly and removes the conflicting color instruction.

## Common questions

<details open>

<summary>Do I need to use parentheses or prompt weights?</summary>

No. Start with a clear natural-language description. Some models support weighting syntax, but the exact format depends on the model.

</details>

<details open>

<summary>Can I use keywords instead of full sentences?</summary>

Yes. Some models work well with short keyword prompts, but sentences can make relationships clearer—for example, which person is wearing a coat or where an object is placed.

</details>

<details open>

<summary>Why does the same prompt produce different images?</summary>

The model and generation settings can affect the result, and some variation is normal. Reusing the same words doesn't always produce the same image.

</details>

<details open>

<summary>How do I change just one detail?</summary>

If you generate again from text, other parts of the image may change too. To preserve the existing image, use a compatible image-editing tool and describe what to change and what should stay the same.

</details>

## Ready to create?

Start with a simple scene, generate an image, and adjust the details that matter most. If text alone can't preserve the pose or identity you need, try a reference-based workflow.

[Create an image in SeaArt AI ↗](https://www.seaart.ai/create/image)