Global Economic Times

SOYNET Launches ‘DABAR-L,’ a Secure On-Premises AI Interpretation Server… Begins Full-Scale Push into the Asian MICE Market: Park Jeong-woo Specialized Reporter / Updated : 2026-04-01 08:57:45

Pangyo-based AI startup targets the Asian market with an AI interpretation appliance that prevents cloud data leaks…

Park Jeong-woo, CEO, SOYNET Co., Ltd. (C) Park Jeong-woo

Seongnam, Gyeonggi-do, March 29, 2026 — SOYNET Co., Ltd., a South Korean AI voice technology specialist recognized as one of Asia’s leading AI startups, has officially launched “DABAR-L.” DABAR-L is a fully on-premises AI simultaneous interpretation and translation server designed for enterprises, public institutions, and MICE (Meetings, Incentives, Conferences, and Exhibitions) operators that cannot expose sensitive conversation content to external cloud infrastructure.

This product launch marks a new strategic shift beyond the cloud-based DABAR platform and is a move to directly address the rapidly growing demand for data-sovereign AI in the Asian conference and government sectors. SOYNET is currently actively recruiting global distribution and co-marketing partners, with a particular focus on the Asian market where the MICE industry is well-developed. Key target industries include international conference organizers, government ministries and agencies, corporate R&D centers, the defense and security sectors, and universities hosting multilingual academic events.

Problems Solved by DABAR-L

As AI interpretation becomes a standard feature at international events, critical vulnerabilities are coming to light. Virtually all major AI interpretation services currently on the market transmit voice data to external clouds in real time. From the perspective of technology companies, manufacturers, government agencies, and research institutions, this poses a serious information security risk. This is because the contents of technical documents, strategic discussions, contract negotiations, and R&D sessions are transmitted in their entirety to external parties and can be used to train third-party AI models.

This data serves as training data for global AI companies to refine their models, effectively allowing competitors to advance their technologies using the companies’ confidential information.

DABAR-L addresses four key issues with existing voice-to-voice interpretation. First, ASR (Automatic Speech Recognition) errors caused by pronunciation, dialects, and background noise; second, the risk of data leakage when voice data is transmitted to cloud servers; third, latency issues that disrupt the flow of conversation; and fourth, contextual errors where humor, idioms, and cultural references are either not translated or lead to misunderstandings. The on-premises architecture fundamentally eliminates privacy and latency issues by completing all processing within the customer’s internal network.

Technical Specifications

DABAR-Ll supports over 30 voice recognition languages using the Chrome Engine and models from Whisper and Voxtral. It also features a machine translation engine capable of translating over 100 languages and a Chrome Engine-based TTS. By combining various pre-processing AI capabilities—such as speech extraction and noise reduction—it performs secure interpretation through secure servers installed at the client’s premises, without relying on the cloud.

The client interface is provided as a web app, eliminating the need for dedicated devices on the participant’s end and allowing access via standard PCs or mobile browsers. Its key differentiating feature is the “Meeting Room.” It provides on-premises n:n multilingual simultaneous interpretation for multi-party environments such as roundtable discussions, executive briefings, meetings with foreign investors, and classroom lessons. Participants set their preferred language upon entering the room, and each speaker’s words are delivered in real time as subtitles or audio in the listener’s language. The system also supports a presenter screen switching feature. Additionally, through personalized voice training, users can register and train the system on misrecognized words or technical terms, thereby continuously improving recognition accuracy.

Proven Track Record

Founded in Pangyo in 2018, SOYNET has garnered recognition for its technological capabilities both domestically and internationally, including being named one of the Slush Global Top 100 companies in Finland, selected as the Best AI Startup in Asia by CIO Advisor, designated as a TIPS (Technology Innovation Promotion Scheme) company, and awarded the Korea Industrial AI Award by the Korea Information and Communications Technology Association.

SOYNET is not a newcomer to the AI interpretation field. Its cloud platform, DABAR, recognizes 37 languages and provides real-time interpretation in 124 languages via subtitles and audio, and is used in both international conferences and small-scale seminars. In January 2025, DABAR was selected as a venture startup product by the Public Procurement Service, and in August 2025, the company officially launched “DABAR Hybrid,” a next-generation platform where AI interpreters, human interpreters, and stenographers collaborate organically within a single system.

SOYNET is currently actively recruiting global distribution and joint marketing partners, with a particular focus on the Asian market, where the MICE industry is well-developed. Key target sectors include international conference organizers, government ministries and agencies, corporate R&D centers, the defense and security sectors, and universities hosting multilingual academic events.

For inquiries regarding Davarel’s distribution partnerships in Asia and global markets:
Park Jeong-woo, CEO, SOYNET Co., Ltd. support@soynet.io https://dabar.soynet.io