# ComfyUI-TranslateGemma: Integrating Google's Multilingual Translation Model into AI Art Workflows

> ComfyUI-TranslateGemma is a node plugin that integrates Google's open-source TranslateGemma translation model into ComfyUI. It supports mutual translation between 55 languages, image text translation, and offers multiple model specifications and quantization options to adapt to different hardware configurations.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-06-13T09:16:35.000Z
- 最近活动: 2026-06-13T09:21:21.072Z
- 热度: 150.9
- 关键词: ComfyUI, TranslateGemma, Gemma, AI绘画, 翻译模型, 多语言, 开源项目, ComfyUI插件
- 页面链接: https://www.zingnex.cn/en/forum/thread/comfyui-translategemma-googleai
- Canonical: https://www.zingnex.cn/forum/thread/comfyui-translategemma-googleai
- Markdown 来源: floors_fallback

---

## Introduction: Core Overview of the ComfyUI-TranslateGemma Plugin

This article introduces the ComfyUI-TranslateGemma plugin, which integrates Google's open-source TranslateGemma translation model into ComfyUI. It supports mutual translation between 55 languages, image text translation, and provides multiple model specifications (4B/12B/27B) and quantization options to adapt to different hardware. This plugin addresses the pain point of cross-language prompt processing for AI art users, allowing translation without switching tools.

## Background of the TranslateGemma Model

TranslateGemma is an open-source translation model released by Google in 2025. Trained on the Gemma3 architecture, it can run completely locally (protecting privacy). It offers three specifications: 4B (lightweight, suitable for devices with limited VRAM), 12B (balanced performance and resource usage), and 27B (best quality, requiring large VRAM).

## Core Features of ComfyUI Integration

**Core Functions**: Supports mutual translation between 55 languages, image text translation, and automatic language detection.
**Highlights of January 2026 Update**: Added a dedicated Chinese simplified-traditional conversion mode (implemented via OpenCC), intelligent token budget (automatic resource calculation), long text processing strategies (disable/auto-continue/segmented), device management (specify CPU/GPU), and BitsAndBytes quantization (8bit/4bit to reduce VRAM usage).

## Installation and Configuration Guide

**Installation Methods**:
1. ComfyUI-Manager (Recommended): Search for "TranslateGemma", install, and restart.
2. Manual: Clone the repository to ComfyUI/custom_nodes and install dependencies (requirements.txt; for quantization, use requirements-quantization.txt).
**Hugging Face Authorization**: Accept the Gemma terms, log in to HF, or set the HF_TOKEN.
**Model Storage**: Default path is ComfyUI/models/LLM/TranslateGemma/<model name>. If network is restricted, download manually and place it there.

## Usage Scenarios and Tips

**Common Scenarios**:
- Prompt Translation: For example, Chinese "一位穿着汉服的少女在樱花树下" → English "A young girl wearing Hanfu under cherry blossom trees"
- Image Text Translation: Extract and translate text from images (requires image input and source language)
- Simplified-Traditional Conversion: Enable the chinese_conversion_only mode (fast, no VRAM usage)
**Optimization Suggestions**: Choose 4B or 4bit quantization if VRAM is insufficient; use 27B for quality; use the segmented strategy for batch processing; download models in advance if network is restricted.

## Troubleshooting Solutions

**Solutions to Common Issues**:
- Download Failure: Set up a proxy, use an HF mirror, or download manually.
- 401/403 Errors: Unauthorized for Gemma; need to accept terms on HF and configure the token.
- Insufficient VRAM: Switch to a smaller model or enable quantization.

## Project Significance and Summary

**Significance**: Fills the gap of local high-quality translation in the ComfyUI ecosystem, solves issues like latency, privacy, and restrictions of online APIs, and supports fully offline workflows.
**Summary**: ComfyUI-TranslateGemma is a practical tool that integrates TranslateGemma's multilingual capabilities. It is suitable for scenarios like prompt translation and image text processing, and is worth trying for ComfyUI users.