KUSART PRODUCT GUIDE · 2026

New to Anima? Start by generating

Start with the basics: choose the model, add a style, write a prompt, and generate. Learn LoRA only after you are comfortable creating images.

Create with Anima See generation steps

Updated July 23, 202612 min readProduct guide

PART 01 · IMAGE GENERATION

Generate with Anima

Make one clear image first, then change the style, prompt, or switches one at a time. This makes every iteration easier to understand.

Choose Kusa-Anima

Open Image Generator and select Kusa-Anima from the model list.

Add an official style

Choose one official style close to your goal. Beginners should start with a single style.

Write prompts and parameters

Set the prompt, aspect ratio, and image count. Use the Tag Library when needed.

Generate and review

Review subject, composition, and style, then change only one variable.

Choose Kusa-Anima — Open Image Generator and select Kusa-Anima from the model list.

Add an official style — Choose one official style close to your goal. Beginners should start with a single style.

Use the Tag Library when you do not know what to write

The Tag Library turns style, character, composition, and lighting tags into visual options. Click a tag to add it to your prompt.

StyleCharacterCompositionLighting

Write prompts and parameters — Set the prompt, aspect ratio, and image count. Use the Tag Library when needed.

Use the Tag Library when you do not know what to write

How to use the two switches

Generation acceleration

Use it for drafts and fast iteration. Fine details may differ slightly from standard generation.

Style enhancement

Strengthens anime styling, detail, and overall consistency. Best for images closer to final.

Confirm the subject, composition, and style direction before refining.

Advanced workflow

LoRA is for experienced creators

LoRA training is not required for beginners. Move into this professional workflow only after you are comfortable generating with Anima and need the same character, style, scene, or pose to appear consistently.

Still exploring images: keep using the base model and official styles.

Professional work that needs a reusable concept: continue to the LoRA training workflow below.

PART 02 · ADVANCED WORKFLOW

Advanced workflow: train an Anima LoRA

This section is for experienced creators who need consistent characters, styles, scenes, or poses. Choose the right preset first, then carefully review your dataset, captions, and preview prompt.

Open model training

Start by choosing the goal, not by tuning parameters

Style

Learn recurring color, linework, light, and composition.

Chibi

Learn small proportions, simple features, and expressive faces.

Scene

Learn a place, spatial layout, or environment mood.

Character

Keep face, hair, colors, and signature accessories recognizable.

Pose

Learn an action, gesture, camera angle, or repeated composition.

01 · Define the goal

Choose a preset and enter the basics

Choose Style, Chibi, Scene, Character, or Pose. Enter the model name, trigger word, and preview prompt used for training samples.

1.3 Enter the name, trigger word, and preview prompt

02 · Prepare the dataset

Upload, crop, and keep every image focused

Upload images or a folder containing images and TXT captions. Open any image to crop it. A focused dataset is easier to train consistently.

2.1 Upload images or a folder with TXT captions

03 · Generate and review captions

Automatic captions are a starting point

Choose classic WD14 or Smart captions. Open each image and remove temporary clothing, action, prop, or background tags you do not want the LoRA to learn.

3.2 Review and edit captions image by image

04 · Train

Confirm the dataset and follow progress

Confirm images, captions, and the preview prompt before starting. The progress page shows preview images so you can see whether the concept is being learned.

4.1 Confirm the setup and start training

4.2 Training progress and preview images

4.3 Preview results from each checkpoint

05 · Test and use

Open the trained model and send it to Image Generator

Training Management lists every project and model file. Open a model to see its trigger word and previews, then click Use to launch Image Generator with the LoRA.

5.1 View all projects in Training Management

Final check before training

Keep one clear concept per dataset.

Use a unique trigger word and include it in the preview prompt.

More tags are not always better. Remove temporary traits.

Compare checkpoint previews before choosing the final version.

PART 03 · LOCAL LORA

Upload a local Anima LoRA

If you already have an Anima LoRA file, you do not need to retrain it. Upload it to My Styles, set its trigger word and weight, then generate.

Add style

Click Add Style in Image Generator.

Open My

Switch to My and click Upload LoRA file in the lower-right corner.

Complete the form

Upload the file, enter a name, trigger word, and weight. Cover and prompt fields are optional.

Apply and generate

Choose it from My. You can combine style, character, pose, and scene LoRAs.

Add style — Click Add Style in Image Generator.

Open My — Switch to My and click Upload LoRA file in the lower-right corner.

Upload form fields

LoRA file: Required
Name: Required; defaults to file name
Trigger word: Required
Cover: Optional
Default strength: Required; default 0.8
Quality prompt: Optional
Negative prompt: Optional

Apply and generate — Choose it from My. You can combine style, character, pose, and scene LoRAs.

The uploaded LoRA appears under My and can be combined with other styles.

Start now

Generate once, then turn a strong idea into a model

Validate the direction with Anima, train a repeatable character, style, or scene as a LoRA, and keep creating.

Create with Anima Train a LoRA

Before publishing or using results commercially, confirm your image rights and review the current Anima model and KusArt platform rules.

Events & Updates Complete Anima Guide