This is the AI Era. Companies are racing to incorporate as many AI features as possible into their devices. While some features may seem like gimmicks, others are intriguing but lack accessibility, and a few are genuinely useful with real potential. Samsung’s approach to AI is through Galaxy AI, which is available on high-end phones and tablets and houses various AI-driven features.
One of these features is the Live Translate During Calls function. This article will cover how to use this feature and share my findings after testing it with my teammate, Kareen Kiwan.
Table of Contents
What Does the Feature Do?
Galaxy AI includes multiple translation features, but this article will focus on using Live Translate specifically during calls. Primarily, this feature aims to overcome language barriers by providing real-time translations for both parties on the call. The translations are performed offline after downloading the necessary language packs. Initially supporting only 13 languages, the feature now extends to languages like Arabic and Turkish. Furthermore, while it began with support for calls via the Samsung dialer, it now also works with popular VoIP services like WhatsApp.
Note: Availability may vary by region.
Enabling and Configuring the Feature
If you have a compatible device, navigate to Settings > Galaxy AI Settings > Live Translation to enable and configure the feature. Here’s a breakdown of the available options:
- Me: Under this section, you can configure the following:
- Language: Set the language you will speak.
- Voice: Choose the voice for translations if multiple voices are available for the selected language (Samsung TTS voices).
- Mute My Voice: Toggle this to mute your voice so only the translation is heard by the other party. If disabled, they will hear both your voice and the translation.
- Other Person: Similar options are available for the other party, allowing you to set their language, voice, and the option to mute their voice.
- Language and Voice Presets for Each Person: This section lets you create presets for specific contacts or phone numbers. For example, if you frequently call one person who speaks Turkish and another who speaks French, you can set Turkish as the default language for one contact and French for the other. This option appears to work only for regular phone calls made through the Samsung dialer; it does not support third-party dialers.
- Live Translate in Other Apps: Enabling this section provides a list of compatible voice-calling apps, along with the overview: “Get real-time translation in supported voice calling apps. During a call, tap ‘Live Translate’ in the quick panel to start translating. When your mic is muted in some apps, translation of your voice may continue, but others on the call will not hear it.” The feature supports a number of popular apps. On my device, supported installed apps include WhatsApp, Telegram, Meet, and Instagram. Each app in the list has a switch to enable or disable live translation.
Alternatively, you can access Live Translation settings via the Samsung Phone app by tapping More Options > Settings > Call Assist to reach the same settings page.
Using the Feature
In Traditional Phone Calls
To enable Live Translation during a traditional phone call using the built-in Samsung dialer, initiate the call, and locate the Call Assist button, which appears in the middle of the screen above the video call button. Tap it, and you’ll see the Live Translation option.
In VoIP Service Calls
For supported VoIP apps, you can access Live Translation from your device’s Quick Settings. To open Quick Settings, swipe down from the notification shade with two fingers, or, if you use a screen reader, you can assign a gesture to open it directly. It’s also accessible from the main menu of the Jieshuo screen reader or from the TalkBack menu, though Quick Settings may not be included by default in TalkBack.
In Quick Settings, locate the Live Translation option. On my device, it’s the first item on the second page, but you can rearrange Quick Settings items if desired (not covered in this article).
When activating the feature for the first time, you must grant three permissions: Microphone, Call Logs, and Contacts. After granting these permissions, you’ll see a prompt saying “Break language barriers” with an OK button. Tap OK to start the service.
Once enabled, an announcement will play for both parties in the languages set in the Live Translation settings. For instance, if your language is set to Turkish, you’ll hear the announcement in Turkish, while the other party will hear it in Arabic if that’s their set language. This announcement also provides information, such as that you’ll see translations on the screen in real-time during the call.
When active, the service listens for either party to speak, translates the spoken text, and plays back the translation. A transcription of the detected text and its translation appears on the screen, navigable and reviewable with the screen reader. An option to quickly switch languages is also available on the screen if needed.
Our Tests
An AI-powered live translation feature is undoubtedly intriguing, but its true value lies in real-world performance. I conducted two WhatsApp test calls with Kareen: I spoke in Turkish, while she spoke in standard Arabic (without any local dialect). We then verified what we had said to assess the accuracy of the translations.
I used my Galaxy S23 and Kareen used her Galaxy A55 that doesn’t have the feature.
A Potential but Not Up to the Hype
Positives: Clear Voice and Quick Accessible Translations
Starting with the positives, both Kareen and I agree on the solid implementation of this feature. The translations were clear, and both of us could easily hear the Turkish and Arabic translations. We appreciate this approach, as either or both parties on a call might understand some of the spoken language or certain terms, making real-time verification of translations valuable in such cases.
The speed of translation was also impressive, with only a short wait time, even for slightly longer sentences.
Moreover, including the option to change languages during the call is a well-thought-out addition.
Another aspect I appreciated was the accessibility of the implementation. During testing, I encountered no accessibility issues. The screen reader, Jieshuo in my case, did not speak over the translation voice, allowing the translations to come through clearly. Additionally, reading the transcription of both the original and translated text was possible through manual screen exploration and swiping gestures, making the feature easy to navigate.
The Quality of the Translation Leaves a Lot to Be Desired
While online translation services show notable progress, offline translation seems to require further development to catch up. We weren’t expecting comparable quality to online AI services, but we did hope for better translations.
Simple phrases, like greetings and weather-related comments (“The weather is nice today and the sun is shining,” with the reply “Here it is raining today”), translated accurately. However, during a second call, where Kareen asked if she could reserve a table at a restaurant for tomorrow, the translation was initially accurate but became problematic as the conversation continued. My confirmation that this was possible and my follow-up question about the time translated correctly. However, her response specifying “10 o’clock for two people” was misinterpreted, rendering my confirmation as “We have an area for the 2nd table at 10,” using the Arabic term for a data table rather than a dining table. When she requested confirmation, “10” was misinterpreted as “12,” and again, the dining table was rendered as a data table. My response appeared as “I confirm your reservation of table 12 tomorrow morning.” This confusion led her to ask if the time was 10 or 12, which I clarified, and this time, “10” was interpreted correctly.
A key observation here is the translation of “table.” In Arabic and Turkish, unlike English, distinct words differentiate a data table from a dining table. However, the translation rendered “table” using the Arabic term for data tables. This error suggests the service may first translate to English, a language that uses the same word for both meanings, before translating the text to Arabic.
In another instance, the translation was so inaccurate that it turned one sentence into “Let’s talk about medical topics” and another into “Let’s talk about kids.”
A significant issue with this service is its lack of contextual awareness despite being an AI feature. Effective language translation requires more than literal accuracy, a problem now largely resolved in advanced online services. Unfortunately, our tests indicated that this service relies heavily on literal translations. For instance, if the AI had some contextual understanding, it likely would have avoided translating the Turkish word for a dining table as a data table in Arabic.
Final Remarks
The in-call live translation was one of the few AI features that caught my attention, especially with its expanded language support and inclusion of VoIP services. Testing revealed a solid implementation, especially since I didn’t encounter accessibility issues. Given that processing happens offline on-device, the response time is acceptable, and the translations worked for basic questions and statements.
However, I wouldn’t consider the feature reliable due to its frequent translation errors and lack of context awareness. Currently, it’s not recommended for important conversations. Translation quality may vary across languages, so testing it, if available, can still be worthwhile, even if not intended for regular use. Live translation serves as a reminder of the progress made in breaking language barriers, with phones now performing real-time translations—a capability that seemed like a far-off dream not long ago. With AI clearly here to stay, following its development in language understanding and translation is essential, as its potential impact on international communication is significant.
Good