improvements in the design through feedback from users. Built upon a layered architecture, it is made up of an

in-built translation engine, text-to-speech and speech-to-text modules, and offline capabilities via downloadable

language packages. Additional important features, including context-aware translation, formality detection, a

correction facility, and feedback loop for user suggestions, ensure adaptability to the linguistic and cultural

diversity of Southeast Asia.

SEA survey was structured for the evaluation of the system, it comprised 100 respondents randomly selected

from three different establishments to assess the system based on seven attributes of quality: functionality,

performance, usability, reliability, security, maintainability, and compatibility. Results showed high marks

across the board ranging from mean scores of 4.07 to 4.22 (Agree), with the biggest scores mostly given to

functionality and compatibility, signifying the system can deliver the essential feature without neglecting

languages, mobile application

Speech cannot co-exist among speakers of different languages. This is what SEA-Talk believes in. And with the

advanced technologies, now came the two ways voice translator, which provides real-time fluent

communication. Just type the text you want to translate in real-time. Snap a picture of the text and get it translated

(Christopher Hill, 2020). Based on these, SEA-Talk also employed AI and ML thus helping you to translate

words in an easy way through voice-to-voice system, which gives the comfort for communicating to others, even

if barrier languages exist. AI now plays a great role in changing the interaction patterns with technology, where

machines think and perform tasks more or less like humans. A major component of AI is a Machine Learning

language barriers found across Southeast Asia (SEA) by real-time voice-to-voice translation. Instant voice

translation is available through the app SEA-Talk, allowing users to speak naturally to be translated dynamically

by the app into the selected language and effective across the area. Simply put, SEA-Talk translates simple-day

conversation sentences. Furthermore, it adopts a participatory approach to improve the quality,

contextualization, and accuracy of translated outputs by permitting users to append more languages or data. This

feature guarantees adaptability that is needed to cope up with Southeast Asia's linguistic diversity while

improving general performance of the system.

Baliber et al. (2021) explored multilingual neural machine translation (NMT) for Philippine languages. The

study employed transformer-based models with zero-shot and pivot translation methods, which improved results

Kannan et al. (2024) employed Hugging Face transformer models, including mBART and No Language Left

Behind (NLLB), to design a multilingual translation system. Their approach prioritized the use of large-scale

training datasets to address challenges in low-resource language processing. The system attained high BLEU

scores and exhibited notable fluency and context management, particularly in languages that are often

underrepresented in existing translation platforms.

Zhang et al. (2023) developed the NPU-MSXF Speech-to-Speech Translation System for the IWSLT 2023

evaluation campaign, using a cascaded architecture that combined automatic speech recognition (ASR), neural

machine translation (NMT), and text-to-speech (TTS). The Transformer-based encoder–decoder model achieved

strong BLEU and METEOR scores across multiple language pairs, demonstrating high efficiency in real-time

multilingual translation.

and cultural adaptation. The study integrated emotion and sentiment recognition to improve translation quality

beyond literal meanings. Their work highlighted the importance of designing systems sensitive to tone,

interface.

Ogundokun et al. (2021) developed an Android-based language converter application to address communication

barriers caused by language differences. The system utilized Google’s real-time translation API and natural

language processing (NLP) with Java to translate major global languages such as English, Spanish, Arabic,

Hindi, French, and Chinese. Designed for accessibility and convenience, the mobile app facilitated effective

communication for users, particularly travelers, by enabling real-time translation and language learning.

enhance user experience.

Kothari et al. (2024) developed a cross-platform real-time translation application that supported both spoken and

written communication. Using artificial intelligence (AI) and natural language processing (NLP), the system

enabled speech-to-speech and text-to-text translation in more than 70 languages, with offline functionality for

Figure 2.1: input-process-output model (IPO)

The figure 2.1 shows the stages that the study needs to take in order for the proponents achieve its desired output,

and that will be the mobile fashion recommender with a smart filtering algorithm in order to sort what the user

likes.

● The input phase consists of several necessary elements required to build and operate the voice translator. The

Artificial Intelligence (AI), which serves as the backbone of the system, by authorizing advanced language

processing and recognition abilities. Translation plays an important function in ensuring that the speech or text

is accurately converted between different languages. Speech/Text input allows users to communicate through

voice or written text, which will be processed for translation. Additionally, requirements such as programming

● The outcome of the system should result in a functional SEA-Talk AI-Powered Voice Translator with the

ability to deliver real-time and accurate translation. The system produces the translation output in the user's

desired Language and enables the content accessible in both Speech and Text formats.

In this chapter, the research approach of the carried-out study was described. First, the population was defined

and the study participants introduced. Second, the procedures employed in designing the instruments and the

collection of the data and third, a description of the statistical treatment employed to analyze the data.

Figure 3.1: Agile Model

In this phase, it refers to the direction of establishing the goals, tasks, and priorities of the process in Agile

researchers to develop intricate User Interface (UI) and User Experience (UX) designs. Figma enabled the

researchers to conceptualize the interactions and movements of every function within the application. Design

involved the development of interactive prototypes that highlighted real-time translation interfaces, language

options, speech-to-text capabilities, and user settings.

In this phase, the actual user interface of the system is implemented from the approved prototype and design by

the researchers. The SEA-Talk application was developed with its own built-in translation engine to improve

language translation accuracy, supported by downloadable language packages to reduce storage burden and

allow flexibility for users. It also incorporated advanced speech synthesis functionality. Each release had a

specific focus on functionalities like enhancing translation accuracy, the speed of translation, and responsiveness

of the mobile app.

performance, reliability, usability, security, maintainability, and compatibility.

efficiency, usability, reliability, security, and portability. The results are presented in tables accompanied by

mean scores and were interpreted based on a Likert scale (1=Strongly Disagree to 5=Strongly Agree) and a

graphical scale for a better result perception. The subsequent discussion draws attention to how these results