Anima Multilingual Prompt Workflow with Gemma Translator

This is a ComfyUI workflow for generating Anima-style images from multilingual prompts.
Using a local Gemma/Ollama translation node, it converts prompts written in Chinese, Korean, Russian, Esperanto, and other languages into natural English prompts optimized for the Anima image model.

This workflow is intended for users who want to write prompts in the language they are most comfortable with while creating clean, natural English prompts for Anima-based models.

Main Features

• Supports multilingual prompt input

• Automatically rewrites prompts into natural English

Included Custom Node

This package includes the following:

ComfyUI-GemmaTranslator

Node display name:

Gemma Any Language to Anima Natural Prompt

Internal node type:

GemmaAnyToAnimaNaturalPrompt

This node sends the input prompt to a local Ollama endpoint and rewrites it into a natural English prompt for image generation.

Requirements

Please install the following before use.

Required local tools

• ComfyUI

• Ollama

• A Gemma model for Ollama

Recommended command:

ollama pull gemma3:4b

Ollama must be running at the following URL:

http://127.0.0.1:11434/api/generate

Required ComfyUI Custom Nodes

Depending on the workflow version, the following custom nodes may be required:

• ComfyUI-GemmaTranslator

• ComfyUI-Impact-Pack

• comfyui-impact-subpack

• comfyui-image-saver

• ComfyUI-Custom-Scripts

The following nodes are used in this workflow:

• GemmaAnyToAnimaNaturalPrompt

• FaceDetailer

• UltralyticsDetectorProvider

• Image Saver Simple

• LoraLoader|pysssss

Models Used

This workflow is designed for an Anima / Qwen-image style setup.

Examples:

• Anima_liquidmix_base_v1.12.safetensors

• qwen_3_06b_base.safetensors

• qwenimageVAE_liquid1087.safetensors

Checkpoint, VAE, Text Encoder, and LoRA can be replaced with compatible files available in your own environment.

How to Use

1. Install the required custom nodes.

2. Install and start Ollama.

3. Download the Gemma model.
   ollama pull gemma3:4b

4. Load the workflow in ComfyUI.

5. Enter your prompt into the Gemma Any Language to Anima Natural Prompt node.

6. Run the workflow.

7. Before being passed to the Positive Prompt Encoder, the prompt will be translated and rewritten into natural English.

Additional Notes

The Gemma translation node does not use an online API.
Prompts are sent to a local Ollama server.

If translation fails, the node will return an error message along with the original prompt.
Please check whether Ollama is running and whether the selected model name is correct.

Default model:

gemma3:4b

Default endpoint:

http://127.0.0.1:11434/api/generate

Notes / Important Information

This workflow requires a local environment setup.
If all required custom nodes and models are not installed, it may not work as-is.

The included Gemma translation node depends on a local Ollama environment.
It does not automatically download the Gemma model or start Ollama for you.