# 4-1 Model

> **In SeaArt, there's no longer a need for local model deployment. You can simply choose online, or even upload and train your own models.**

**Creation Process:** <mark style="background-color:red;">Choose model</mark> - Enter prompts - Adjust relevant parameters - Generate

## Checkpoint

Generally speaking, both Lora and Checkpoint are referred to as models. Therefore, we can call Checkpoint the large model or base model. Compared to Lora, the large model is relatively larger, ranging from 2GB to 7GB, and it determines the <mark style="background-color:yellow;">main style</mark> of the final image.

Same prompts, different Model:

<mark style="background-color:yellow;">Prompts: black hair, princess, frills, standing, short hair</mark>

**Recommended Checkpoint:**

Realistic:

Anime:

## LoRA

LoRA is typically around 100MB in size. It can "fine-tune" the style of images, fix the appearance, posture, etc. On top of overlaying the Checkpoint, adjusting weights can achieve different effects. In the SeaArt, Lora can stack up to 5.

**Character LoRA:** These models specialize in capturing specific character traits, such as appearance, body proportions, and expressions, commonly found in cartoons, video games, or other media. They prove valuable for fan artwork creation, game development, and animation/illustration projects.

**Style LoRA:** These models are tailored to mimic the artistic style of specific artists or artistic movements. They excel in transforming reference images into the desired aesthetic, making them useful for stylization purposes.

**Clothing LoRA:** Similar to Style LoRA, Clothing LoRA models focus on specific clothing styles or fashion aesthetics. They are adept at generating images with a particular clothing aesthetic, offering applications in fashion design and digital styling.

Different Lora with the same Checkpoint and prompt:

Prompt: A girl, frontal view, close-up, sitting on a plush chair, exquisite face, full of confidence, luxurious atmosphere, top quality, 8k.

Add multiple Loras:

After adding the Team Rocket Uniform, not only did it increase the weight of Lora, but it also added related prompts for Lora: <mark style="background-color:yellow;">"Team Rocket uniform," "red letter R," "white skirt," "white short top," "black long sleeves," "black elbow gloves,"</mark> making the effect of Lora more pronounced.

> \
> [Previous4-Parameters](https://app.gitbook.com/o/SXzpAogddQhmphNiXJbH/s/igAtVLBrlaI8jVruJfC8/~/edit/~/changes/212/4-parameters)[Next4-2 Mode](https://app.gitbook.com/o/SXzpAogddQhmphNiXJbH/s/igAtVLBrlaI8jVruJfC8/~/edit/~/changes/212/4-parameters/4-2-mode)SULast modified 1yr agoOn this page[Checkpoint](https://app.gitbook.com/o/SXzpAogddQhmphNiXJbH/s/igAtVLBrlaI8jVruJfC8/~/edit/~/changes/212/4-parameters/4-1-model#checkpoint)[LoR](https://app.gitbook.com/o/SXzpAogddQhmphNiXJbH/s/igAtVLBrlaI8jVruJfC8/~/edit/~/changes/212/4-parameters/4-1-model#lora)If the effect is not significant after adding Lora, we can try:&#x20;
>
> I. Adjusting the weight of Lora&#x20;
>
> II. Entering corresponding prompts

Compared to Checkpoint, Lora has shorter training time and higher "flexibility," and can exert good effects on controlling images.

<mark style="color:red;">Attention!!!</mark>

I. SD1.5 and SDXL's Lora cannot be used together.

II. Most Lora require trigger words. Including these words in the prompts can emphasize the unique subject or style provided by the Lora. These trigger words are not always necessary, but it is recommended to include them in the prompts to ensure stable image generation.

**Different Lora Effects:**

1. **Detail Tweaker LoRA:** The higher the weight, the more details.

<figure><img src="/files/2n3SGgHcf2SMzU15bNmU" alt="AI Image Generation Model - Detail Tweaker LoRA" width="563"><figcaption></figcaption></figure>

2. **epi\_noiseoffset**: Increases contrast.
3. **Makima (Chainsaw Man) LoRA**: Fixes the character's image.
4. **ankymoore illustration**: Modifies the art style.

<mark style="color:red;">\*</mark>*Lora has many functions, trying different Loras may bring more surprises!*


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.seaart.ai/guide-1/4-parameters/4-1-model.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.