MyText.ai

Context is Important

While other translation services only focus on the words in a text, our AI agent considers the context in which the text appears. It knows, for example, when a text is actually a button label, a page title, or a paragraph. This allows us to translate more accurately and consistently, ensuring that the translated text is appropriate for the context in which it will be used.

Right In the IDE

To understand the text context, we use an extension (VS Code, Cursor, and Visual Studio) or plugin (Android Studio, IntelliJ and Rider) that you install in your IDE, which allows our agent to locally read your project’s source code. Note that your source code is not sent to our servers; only small pieces of contextual information that do not constitute important proprietary information are used.

The AI Does the Heavy Lifting

Another issue is that most translation services don’t offer an easy way to integrate translation into your workflow. Through the extension/plugin, MyText.ai can implement and configure translation and localization packages, select texts for translation, and translate your source code without you having to leave your IDE, reducing or eliminating the need for all of this to be done manually.

If you’ve used other translation services, you know this can be a time-consuming and complicated process, taking up the time of several team members, depending on the project size. With MyText.ai, in most cases both the installation of translation packages and the translation itself will be done with zero or very low effort. The cost saved by this reduction or elimination of labor can be significant, allowing translation projects that were previously unfeasible to be carried out.

Swarm of OpenAI Translators

Last but not least, translations done by MyText.ai use state-of-the-art models from OpenAI. These advanced models can understand text deeply, taking into account the culture of the country or region of the language, and have an understanding of technical jargon from almost all areas of knowledge.

For each text to be translated, MyText.ai creates a swarm of independent AIs, some GPT-4 and others GPT4o, which first translate the text but then converse among themselves to discuss the quality of the translation, try alternative translations and discuss which is best, translate the text back to the default language to compare with the original text, discuss again, and finally, translate the text to the target language.

A simple translation of a paragraph can generate pages of discussion among these AIs, forcing the model to “think” at length about each word. This ensures that the translation is as accurate as possible, and that the translated text is as close as possible to the intention of the original text.

For Indies, Startups, and the Enterprise

If you’ve used other translation services before, you might have noticed their process can be slow and complex, often requiring involvement from multiple members of your team. These translation services tend to focus on larger clients who, because of their own high labor costs in supporting the translation effort, are less sensitive to premium fees for the translation service itself. Essentially, these translation services understand that their larger clients are already managing significant internal expenses, making the additional high costs charged by the translation service seem relatively minor in comparison.

With MyText.ai, both preparing your project and importing the translated texts require little to no effort, thanks to our AI agent. The significant time and cost savings from reducing or eliminating labor can make previously unfeasible translation projects viable. This is where most of the savings from MyText.ai originate. But also, these reduced costs attract customers of all sizes, from solo entrepreneurs to small and large companies, allowing us to maintain a larger customer base while operating with smaller margins per client.

AI Translations are the Future

Translating text through AI is much faster and cheaper than using human translators. Although other translation services also allow you to use AI, there’s an incentive for these services to not be very good so that you would prefer human translators. This is because human translators charge much more, and therefore, the translation service can earn more. If AI can translate quickly and accurately, clients might prefer AI, and human translators could lose business. This creates an incentive for translation services not to invest much in AI, and to keep its quality mediocre. This is the case with most AI-using translation services.

MyText.ai, however, has already achieved quality equal to or better than human translators for the vast majority of languages, using OpenAI’s GPT-4 and 4o models, and our proprietary AI swarm technology. Very soon, as AI base models improve, AI will soon surpass human translators for all languages.

Texts Committed to your Own Repositories

Lastly, several other translation services store the translated text on their servers, or even serve the translated text real-time, online, turning them effectively into cloud services with much higher maintenance and operation costs. MyText.ai, in contrast, sends translations to you, allowing you to commit them to your own repository. This reduces maintenance and operation costs, allowing us to offer lower prices.

MyText.ai can translate to most existing languages, dialects, regional variations, playful variations, and even invented or dead languages.

Example:

Language translations: English, Spanish, French, etc.
Dialect translations: Standard German, Bavarian, Swabian, etc.
Regional variations: Brazilian Portuguese, European Portuguese, European Spanish, Mexican Spanish, etc.
Playful variations: Cockney, Verlan, Pig Latin, Lunfardo, Caló, etc.
Invented or dead languages: Esperanto, Latin, etc.

How many languages should I choose?

In theory you could translate your project into thousands of languages and variations, but in practice, to control costs, you should consider prioritizing the most widely spoken or relevant languages for your target audience. Focus on languages that will have the greatest impact, or those that align closely with your business goals, geographic markets, and user demographics. Additionally, you might want to start with a smaller selection of languages and expand gradually based on the feedback and demand from your audience.

To start, and learn how to use MyText.ai you can begin with a single language. If your project is in English, translate it to Spanish. If your project is not in English, translate it to English. This will allow you to see how the translation process works and how the translated text is integrated back into your project.

Then, you can expand to more languages. If you have a truly global focus, this is a recommended list of 40 languages: English, Chinese (Simplified), Spanish, Arabic, Hindi, Portuguese (Brazil), Russian, Japanese, German, French, Indonesian, Korean, Italian, Turkish, Vietnamese, Thai, Portuguese (Portugal), Dutch, Polish, Bengali, Filipino (Tagalog), Urdu, Malay, Ukrainian, Tamil, Hebrew, Swahili, Romanian, Greek, Czech, Hungarian, Swedish, Danish, Finnish, Norwegian, Serbian, Bulgarian, Slovak, Croatian.

Suported content types include:

Games

Console games
PC games
Mobile games
Game images

Apps

Mobile apps
Web apps
Desktop apps

Content rich

Websites
Documentation
Marketing materials

Note we are currently refining our approach to translating images. One method under consideration involves using Optical Character Recognition (OCR) to extract texts from images, which are then translated. The original text is masked, and a diffusion model is employed to seamlessly fill in the background. The translated text is then superimposed onto this clean slate with the same fonts and colors. Alternatively, we are exploring integrating MyText.ai directly into the design process, enabling the creation of multilingual designs from the outset.

Translations of widely spoken languages will generally be of higher quality than those of average professional human translators. For less commonly used languages, quality may depend on factors like the number of speakers and the language’s presence on the internet and in books. In most cases, the translation quality will be comparable to that of an average professional human translator.

Note MyText.ai takes into consideration not only the language, but also the culture of the country or region associated with that language. The AI has a deep understanding of cultural nuances and technical jargon across various fields of knowledge, which human translators do not allways possess.

Finally, you may have noticed that websites translated by humans sometimes display a mix of English and another language. For example, take this actual screenshot of a website intended to be in Portuguese, yet it shows both Portuguese and English:

Probably, the text above on the left side is more recent than the text on the right. Human translation processes are periodic and can be slow, so the newer content hasn’t been translated yet, leading the site to retain the original English. MyText.ai addresses this by offering “continuous translations,” which may be performed seconds after new text is added.

Frequently Asked Questions