In the midst of the development of artificial intelligence, the question floats in the air as to whether we already have the technology to fulfill an old dream: that we can talk to anyone, regardless of the language they speak.
Digital translation services work reasonably well. In 2024, are they able to interpret what we hear in another language easily, reliably and quickly? On TikTok we see that there are many videos of people using Bluetooth headphones to have conversations with others who do not speak their language.
In all of them it seems that the technology works well. We wanted to check to what extent a very popular model among some TikTokers, the Timekettle headphones, is reliable. We wanted to put them to the test, as they seem to be among the best systems that exist on the consumer market for instant translations.
For several days we have tested the effectiveness of the Timekettle WT2 translator headphones and a Fluentalk T1 Mini translator. The headphones, which work through an application for mobile phones, according to the manufacturer, allow simultaneous two-way interpretation, something that on paper seems quite ambitious. In addition, the manufacturer assures that they have a translation efficiency of up to 95%.
The company has created a translation engine that uses the technology of other engines, such as those from Google, Microsoft, DeeL, or iFlytek. And what seems most important: Timekettle has 15 servers spread around the world. We assume that this is what accounts for the good latency of the service, as it is capable of providing translations with hardly any waits.
We carried out our test with an expert. María José Hernandez Sánchez, English translator and interpreter. We have a conversation with her in which this journalist speaks in Spanish and she speaks in English. And we do it in the noisy environment of a bar. A tough scenario for any translation system.
We have two types of talk, one simple and others more complex. Something that is immediately noticeable is that the translation is noticeably faster than if we use the most popular translation applications that allow a conversation mode, such as Google Translate.
Timekettle has clearly achieved good levels of latency, the time that elapses between the phrase spoken by one of the two interlocutors and the response produced by the device. This is extremely important in order to maintain a conversation. Everyone who has used translators like those from Google or Microsoft has encountered the problem that these applications sometimes take too long to respond.
The next test we do is with a Chinese small business owner. With it it is much more unusual to try these headphones. In this case the environment is much quieter. It is surprising to hear the translation of a conversation from Chinese to Spanish live. We even talk about politics, a thorny topic that is not easy for a translator. The result is quite satisfactory and we found that the ambient noise of the previous environment, with music playing, was excessive.
In both tests this journalist uses one headset and the other person the other. The product comes with disposable protectors to maintain hygiene. It is the most ideal way to avoid background sounds, since the microphone is specially adapted to capture the voice. Something that is useful if we want, for example, to listen to the translation of a voice that comes out of a speaker, for example.
The speed of the translation is high if we read on the screen what the other person says in the other language. Although the conversion of text to speech is somewhat slower. Until the person finishes speaking, you don't hear what he or she has said. But little by little, as you chat, you decide to leave the phone in your pocket and let the conversation flow through the headphones.
The quality of translation, as with other services, has to improve. Although for someone who does not have a good level of understanding of English, such as this journalist, it is important to understand what another person is saying in that language, even approximately.
It is perceived that some grammatical constructions make little sense or that some words are not correctly translated. In fact, by the standards of an expert in this field, like Hernández, the results are far from what a human interpreter can achieve.
This expert alerts us to something important. One thing is consecutive translation, which in her opinion is what the technologies she has tested do, and another is simultaneous translation. This requires greater speed and so far only professionals can do it effectively.
But the response speed of these translation headphones is an important innovation. We found that the Fluentalk T1 Mini portable translator is somewhat slower, but it has in its favor that it does not require a phone to work.
Fluentalk T1 Mini can be used with a SIM card that allows you to have data worldwide for free for a year (it can be recharged for about $15 a month or $45 for a year). You can also connect to Wi-Fi networks. Its best quality is its small size and that it incorporates a small camera to translate texts.
Its use makes sense when we do not want to drain the phone's battery. Both the Timekettle WT2 headsets and the Fluentalk T1 Mini translator can download language packs for offline translations.
Translation and interpretation technologies will surely improve a lot in the coming years. The translator with whom we carried out this test explains to us that a large part of her work already consists of polishing translations generated with AI.
But if we don't want to wait, the Timekettle WT2 headphones - which cost 350 euros with a package that allows you to download several languages - are interesting so as not to feel lost in environments with unknown languages. Of course, it is better not to have too high expectations. The Tower of Babel still stands. Although the learning curve is small, it is advisable to learn to use this device fluently so that it does not bring us unpleasant surprises in improvised environments.
Beyond the capabilities of Timekettle's products, it seems clear that technology exists to accurately interpret languages in real time. But increasing translation quality and speed depends largely on deploying a powerful server infrastructure. As Microsoft has done so that Open AI has been able to deploy ChatGPT. And that is expensive, very expensive.
