Translation AI

What is AI translation?

Huìyì AI, also known as Huìyì·Contrastive Translation, is a multilingual translation platform operated by Wuhan Nightingale Technology Co., Ltd. The product is based on a browser plugin, and it also offers services for translating online text, PDF files, Word documents, images, video subtitles, as well as providing API solutions for document translation.

Its main advantage is that it allows the original text and its translation to be displayed simultaneously on a webpage or in a document, thereby reducing the need to switch back and forth between the translated page and the original page. Individuals can use it for reading and learning, while professional users and teams can take advantage of glossaries, bulk file processing, larger capacity limits, and API access.

Main functions

Bilingual translation of web pages

• After installing the plugin, open a foreign-language webpage and use the floating menu on the side to initiate full-text translation; the original text and its translation can be viewed side by side, making it suitable for news articles, technical documents, product pages, and overseas materials.
• Word translation allows you to see the translation immediately after selecting a word or phrase, while hover translation enables you to use a shortcut key and move the mouse over a paragraph – it’s suitable for translating only specific parts of text.
• The translation function for input fields can directly process the text in web forms, while search translation allows the translated results to be filled back into the search box, facilitating searches across different languages.

Online text and multilingual translation

• Users can paste words, sentences, or paragraphs for translation; the system can automatically detect the source language, or it is possible to specify the input and target languages manually.
• The public page states that it supports translation between 54 different languages, making it suitable for temporarily translating text outside of a browser; the specific language combinations and available models are indicated in the current translation panel.
• Different models are better suited for different tasks: basic models are designed to understand content, while advanced and top-tier models are used for more complex contexts and professional texts.

PDF translation and ChatPDF

• PDF segmental translation presents the content in both languages paragraph by paragraph, facilitating reading section by section; lossless translation focuses on preserving the original document’s layout, image positions, and reading order.
• Scan translation is intended for PDF files from which text cannot be extracted directly; it is first necessary to identify the content on the pages, and the quality of this identification depends on the clarity of the scan, as well as the complexity of the font and layout.
• ChatPDF allows users to ask questions, obtain explanations, and get summaries regarding uploaded documents; it is useful for quickly understanding papers, contracts, reports, and manuals. However, it is still necessary to check the key terms and data in the original documents.
• The public trial page indicates that ordinary PDF files should not exceed 100 MB in size; the paid plans increase the limit on the number of pages per file as well as its size, with specific details available in the member dashboard.

Translation of Word, images, and videos

• Word Translation accepts document uploads; once the translation is complete, the resulting file can be downloaded. It is suitable for reports, manuscripts, and office documents. The maximum size of a single file varies depending on the package chosen.
• Image translation can handle local images, screenshots, and web images; it is suitable for product posters, explanatory images, and content for social media. The number of images that can be processed each month is limited by the member level.
• Video translation is used for translating subtitles on platforms such as YouTube and Bilibili, thereby offering a bilingual viewing experience; the presence of existing subtitles and the structure of the page affect its usability.
• Web pages and videos provide summaries that condense long texts into key points; the bookmarking function allows saving these summaries, making them useful for creating personal reading notes.

Learning modes and glossary

• The learning mode highlights new words on the webpage, and provides word lists for tests such as CET-4, CET-6, IELTS, and TOEFL, as well as a function for reviewing words.
• AI analysis can explain the selected words or sentences, helping to understand the context, grammar, and expression, rather than just providing a literal translation.
• A custom glossary is used to establish fixed translations for names of people, products, and industry terms; it supports synchronization between Trados terminology bases across multiple devices, and is suitable for corporate and professional translation workflows.

Supported platforms and installation methods

The version of the plugin indicated on the official website is v2.2.0, with an update date of April 27, 2026. No separate native iOS or Android applications have been identified; the capabilities available on mobile devices should not be confused with those of browser plugins.

Guide to Using Translation AI

1. Depending on the browser, choose the app store to install the plugin or install it manually; after installation, verify the extension icon and the permissions for the webpage.
2. Register or log in to your account, then select the desired translation quality and language direction for the task in the model and target language settings.
3. Enable full-text translation while reading web pages; use word highlighting or hold the shortcut key for hover translation when only a partial portion needs to be viewed.
4. When processing documents, choose PDF, Word, or image tools; before uploading, check the format, size, number of pages, confidentiality level, and third-party rights.
5. For professional projects, a glossary is created first; then, a small number of pages are used to test names, terms, tables, formulas, and formatting, before proceeding with bulk processing.
6. After the translation is completed, download it and conduct a manual proofread, paying special attention to numbers, negative words, legal responsibilities, medical terms, citations, and chart labels.

Member prices and quotas

Currently, individual and team memberships are displayed on an annual subscription basis, with the equivalent monthly price also shown on the page. The table below shows the public annual price at the time of verification; promotional offers, bonuses, and actual purchase prices may vary, so it is necessary to refer to the settlement page when making a purchase.

Differences between documents and team permissions

The membership quota is updated on a monthly basis and is subject to specific rules regarding various functions; the availability of basic, advanced, and top-tier models varies. The public pages do not provide clear information regarding the procedures for refunding membership fees, so it is necessary to verify the terms related to renewal, refunds, quota expiration, and invoices before making a payment.

Supported translation models

The translation models are divided into basic, advanced, and top-tier categories; the current page lists models such as DeepL, DeepSeek, DouBao, Qwen, GLM, GPT, Claude, and Gemini. The models and their categories may change as services are upgraded, so no particular model can be considered a permanent option.

• The basic version is suitable for reading web pages and consuming entertainment content on a daily basis, while the paid version allows unlimited use of certain core functions.
• Advanced models are used for translation, summarization, explanation, and chatting; Max, Ultra, and Elite offer different numbers of monthly tokens.
• The top-tier models are designed for handling complex texts and professional semantics; Ultra and Elite come with corresponding monthly quotas, while Max is not available at the moment.

Document translation API

Huiyi offers an API V1 for translating documents, intended for developers and businesses; the quota associated with this API is independent of that of web members. At present, only lossless translation of PDF files is available, while support is not provided for scanned documents, segmented translation, Word files, images, or general text.

• Enter the URL of the PDF file that can be downloaded by the server, and specify the source language, target language, model, output format, and any optional callback.
• Tasks are executed asynchronously, and the output can be a PDF containing only the translated text, or a PDF with the original pages alongside their corresponding translated pages; selecting both formats for the same task does not result in double charging per page.
• Once the access application is approved, the necessary qualifications are granted, and the user can create multiple API Keys; these keys share the same quota for processing PDF pages within the account.
• During development and testing, 100 pages are provided free of charge for a period of 30 days; the key itself does not incur any cost – actual usage is covered by the subscription packages available on the purchase page.
• The system first determines the number of pages in the source PDF and reserves the necessary capacity; once the result is generated successfully, that capacity is deducted. The reserved capacity is released in cases of failed downloads, unsupported file formats, parsing errors, or failures in system processing.

API integration process

1. Submit an application for API integration for document translation in the account backend, and wait for the administrator’s approval.
2. After approval, an API Key is created, and the key is stored solely in the secure configurations on the server side.
3. Call the quota inquiry interface to determine the remaining number of pages, and then submit the URL for downloading the PDF along with the target language to create the task.
4. The completion result is obtained through task status queries or callbacks, and a temporary download address is provided upon success.
5. Business-side controls are established for callback verification, request timeouts, retry on failure, temporary file cleanup, and quota monitoring.

API price

API, SDK, and open-source status

Privacy and data security

• Registration may involve handling WeChat profile pictures and nicknames, phone numbers, open identifiers, or email addresses; companies should first inform their employees and obtain their necessary consent.
• Document translation involves submitting Office documents, PDFs, images, or other content that needs to be processed to the server; these files may contain personal information and business-related data.
• The platform may also handle device models, system versions, device identifiers, IP addresses, location data, access and search logs, orders, payment information, and customer service records.
• To carry out translation, infrastructure management, payment processing, customer service, and analysis, the necessary data may be provided to affiliated parties or authorized partners.
• Data generated as a result of operations within China are, in principle, stored within China. The retention period is determined based on the purpose of the service and relevant legal requirements, and such data is deleted or anonymized once that period has elapsed.
• Users can view, correct, and delete their personal information, revoke consent, or stop using it; requests regarding certain rights are usually processed within 30 days, while responses to privacy complaints are generally given within 15 days.

Copyright and commercial usage notes

• Users must have the right to upload documents, images, and subtitles; translating content that is protected by copyright does not automatically grant the right to distribute it publicly or resell it.
• The service agreement grants users rights to use, on a non-exclusive and free basis, the content they upload, edit, and publish; such rights may also include the possibility of sub-licensing. Confidential documents and restricted materials should be uploaded with caution.
• The platform’s web pages, plugins, software, trademarks, and code belong to the operator or the respective rights holders; they may not be reverse-engineered, copied, resold, or used for illegal activities without permission.
• Machine translation may misinterpret terms, negative expressions, and legal responsibilities; contracts, medical texts, patents, and published materials should be reviewed by professionals.

Suitable for users and scenarios

• Students and researchers: Read foreign-language papers and web materials, using word highlighting and ChatPDF to aid comprehension.
• Cross-border e-commerce and operations: Translating product pages, images, overseas reviews, video subtitles, and search keywords.
• Workplace users: Handle Word, PDF, email documents, and technical documents from overseas, using a glossary to ensure consistent terminology.
• Professional translation team: batch processing of large files, management of terminology, use of higher model capacity, and collaboration across multiple devices.
• Developers and enterprises: Integrate lossless PDF translation into document systems, knowledge bases, or business processes.

Advantages and limitations

Main advantages

• Subtitles for web pages, documents, images, and videos are fairly comprehensive, and the browsing path for reading is logical.
• Multiple models, glossaries, learning modes, and batch files are all available, enabling use ranging from personal reading to professional workflows.
• The document API specifies limits for testing, packages with a fixed number of pages, rules for handling failures and returning results, as well as asynchronous callback mechanisms, all of which facilitate the assessment of integration costs.

Usage restrictions

• Memberships and APIs have two separate quota systems; purchasing a membership does not allow for direct deduction of API usage credits.
• The limits on models, tokens, files, retention periods, and annual payment discounts will be adjusted; for long-term projects, the account terms and the contract shall prevail.
• Plugins require web access rights, and the security policies, dynamic structure, or subtitle settings of certain websites may prevent their functionality from being utilized.
• The files will be uploaded for processing; sensitive documents must first have their sensitive information removed, and it is necessary to confirm whether the organization permits the use of external translation services.

Frequently Asked Questions

Can the translation AI be used for free?

You can register for a free account to experience the basic features, but the specific number of free models, tokens, and document quotas are determined by the current account. For frequent use, top-tier models, and processing of large files, a paid plan is required.

What types of files and content can be translated?

It supports scenarios such as web pages, text, PDF, Word, images, and video subtitles. The format requirements, number of pages, and size limits for different file types vary depending on the package chosen.

Are the translation member benefits and API quotas applicable to everyone?

It is not universal. Web members are managed based on model tokens, the number of pages in the document, and various functional privileges, while API access requires separate purchases for packages specifying the number of PDF pages.

Does the translation API support Word files or scanned documents?

Currently, V1 supports lossless translation of PDF files only; it can output either just the translated text or a side-by-side comparison of the original and translated pages. Scanned documents, segmented text, Word files, images, and general text are not yet supported through this API.

How long are translated files stored?

The member page indicates that files and records for the Max, Ultra, and Elite tiers are retained for 1 month, 6 months, and 12 months respectively. Other records related to privacy may be kept depending on the purpose of the service and the legal requirements.

Is Huiyi AI open-source software?

No. No official open-source repository or open-source license corresponding to the product in question was found; moreover, the presence of browser extensions and public APIs does not imply that the source code of the product is available publicly.

Summary

Translation AI is suitable for individuals and teams who need to read content in two languages on websites, translate documents, process images and subtitles, as well as manage specialized terminology. When choosing a package, it is necessary to consider factors such as the number of model tokens, the number of PDF pages, the file size limit, the retention period, and the number of devices; for integration purposes, it is also important to evaluate the number of API calls allowed, file security measures, and the process for manual proofreading.