Image Translation

Introduction

The Image Translation is specially designed for e-commerce images, supporting dozens of language pairs. It accurately handles complex text layouts within images and ensures that the translated content matches the image content. This helps e-commerce platforms and developers easily achieve multilingual conversion and presentation of image content.

We offer two image translation API products: the Image Translation Standard Version and the Image Translation Pro Version, based on Multimodal Large Language Model(MLLM),which enhances translation accuracy and quality through advanced image understanding.

A comprehensive comparison between Aidge's image translation product and competitors revealed that Aidge offers 15% to 20% superior image processing, addressing challenges like color block masking and layout issues. In multilingual translation quality, the Image Translation Standard API outperforms competitors by 10% to 20%. The Pro Version, utilizing a multimodal large model, further improves translation quality by an additional 10% to 15%.

Original Image Input	Image Translation Standard Version Output	MLLM Image Translation Pro Version Output	Description
			Text：160° Electric Recline for Deep Relaxation Image Translation Standard Version Output：160 ° The electric back lies deep into it MLLM Image Translation Pro Version Output：160° Electric Recline for Deep Relaxation MLLM Image Translation Pro Version provides fluent translations consistent with standard English marketing and product description language.
			Text：Brand under EuroComfort Group/69 Years of Experience Image Translation Standard Version Output：Brand owned by german oshine group/69 precipitation MLLM Image Translation Pro Version Output：Brand under EuroComfort Group/69 Years of Experience MLLM Image Translation Pro Version translates names and abbreviations such as "EuroComfort Group" and "Y" on images with English information and brand names.

For more information about the MLLM Image Translation Pro Version, please refer to: https://github.com/AIDC-AI/Ovis

Key Features

The image translation is specifically designed for various e-commerce businesses, facilitating cross-platform e-commerce image creation. It quickly converts main product images and detailed descriptions into languages understood by consumers in different regions, optimizing product presentation and sales across multiple e-commerce platforms.

Key Advantages	Description
High-definition image processing	Capable of handling images up to 4000x4000 pixels while maintaining clarity and quality.
Multiple languages	Supports multiple languages, covering the languages of over a dozen countries and regions, with translation capabilities specifically optimized for e-commerce terminology.
Distinguishes between natural text and artificial text on products, preserving the natural text on the main body of the product.	Differentiates natural product text from artificial text, preserving natural text during translation with product body recognition, ideal for e-commerce images. As shown in the example image, the text in the red box represents the text on the main subject, while the text in the blue box represents text not on the main subject. Input： Output:
Preserves brand name	When brand name protection is enabled, the system can identify artificially added brand names and prevent their translation. In the example image, "懒角落" is the brand name. With protection enabled, it will remain untranslated. Input：Output：
Erases thoroughly, ensuring no residue or color blocks remain.	Input：Output：
Replicates the font style, size, and alignment perfectly .	Input：Output：
The translated Arabic text is arranged from right to left, in accordance with the linguistic characteristics.	Input：Output：
Supports the restoration of text within complex tables after translation.	Input：Output：

Supported Language Pairs

The image translation product supports the language pairs listed in the table below for image translation. We will continue adding more language pairs.

If you have specific needs, please contact us via email, and we will get back to you as soon as possible. Our email address is: [e-mail].

The standard version of the image translation supports a wider range of languages, with 3 source languages including Chinese, English, Turkish, and can translate into over 20 target languages such as Chinese, English, Arabic, Spanish, and French. Currently, 69 language pairs are supported. Among these, language pairs like English to Arabic, and English to European languages are our strengths.

Note: Translation where the source and target languages are the same is not supported (e.g., if the source language is English and the target language is English, the final result will return the original input image).

Source language

Target language

●Chinese (Simplified) ●English ●Turkish

●Chinese ●Arabic ●Bengali ●German ●Spanish ●French ●Hebrew ●Indonesia ●Italian ●Japanese ●Malaysia ●Dutch ●Polish ●Portuguese (Portugal) ●Russian ●Thai ●Korean ●Turkish ●Ukrainian ●Urdu ●Vietnamese ●Kazakh (only Chinese source language is supported) ●Chinese (Traditional)（only Chinese and English as source languages are supported） ●Czech (only English and Turkish as source languages are supported) ●Greek (only English and Turkish as source languages are supported) ●Hungarian (only English and Turkish as source languages are supported )

The Image Translation Pro version (MLLM) currently supports twelve optimized language pairs. The translation quality for each language pair has improved by 10% to 15% compared to the standard version of image translation, and is rapidly expanding.

Source language	Target language
●Chinese (Simplified)	●English ●Japanese ●Russian ●Thai
●English	●Spanish ●French ●German ●Korean ●Portuguese (Portugal) ●Japanese

Pricing

To use the API, you are required to choose and purchase an API resource pack from us on a subscription basis.

• Each resource pack is valid for one calendar year upon successful purchase, and enables you to access the API up to the number of requests specified in the pack. No refunds can be provided.

• If you need to purchase more QPS due to business requirements, please contact us via Discord or email us ([e-mail]).

• Resource packs cannot be used across different products. For example, if you need to use both product text translation and image translation, you must purchase separate resource packs for each.

• The Image Translation Standard Version and MLLM Image Translation Pro Version are priced the same.

The prices are as follows:

Capacity	Price (USD)	Unit Price(USD)
1,000 images	22	$0.022 /image
10,000 images	150	$0.015 /image
100,000 images	1000	$0.010 /image

Quick Start

For instructions on calling the API Submit of the Image Translation Standard Version,please refer to Image Translation Standard Version API Referance. For the API Result Query, please also refer to to document. For the API Submit of the Image Translation Pro Version,please refer to Image Translation Pro Version Result API Call Description. For the API Result Query, please also refer to the document.

FAQ

1. Does it support automatic recognition of images in different languages?

Currently, the image translation only translates the specified input language and does not support automatic recognition of multiple languages.

For example, if an image contains both Chinese and English, and the input source language is set to Chinese, only the Chinese text will be translated while the English text remains untouched. Conversely, if the input source language is set to English, only the English text will be translated while the Chinese text remains untouched.

2. The image translation result has imperfections, and how should I make adjustments?

If your image is not an e-commerce image or does not contain a product main body, or if you encounter missing translations, you can try setting the input parameter to translate the main body text of the product and try again.