SoyNet pushes hybrid and on-premise AI interpretation for secure events

SoyNet is expanding its DABAR real-time interpretation platform into the global MICE, B2B and B2G markets with hybrid and closed-network deployments. The move targets organizations that want lower translation costs, stronger data privacy and live multilingual support without relying entirely on third-party APIs.

Why it matters: - SoyNet is targeting a pain point for conferences, governments and universities: real-time interpretation that is fast, cheaper and less exposed to external data leaks. - The company is pitching private, on-premise and hybrid systems as an alternative to cloud-only tools that send sensitive audio and text through outside APIs. - The approach is aimed at large events, multilingual institutions and security-sensitive users that need interpretation at scale.

What happened: - SoyNet, founded in 2018 in South Korea, is moving its AI-powered interpretation service DABAR deeper into the global MICE market and into B2B and B2G use cases. - The company is based in Pangyo, Gyeonggi-do, and is preparing a funding round of about 1 billion KRW to expand server capacity and sales. - CEO Jung-Woo Park discussed the company’s interpretation ecosystem, security products and overseas expansion in an interview at the Startup Campus in Pangyo Techno Valley.

The details: - DABAR recognizes speech in 37 languages and produces real-time translations in 124 languages. - SoyNet built DABAR on its own Speech-to-Text, Machine Translation and Text-to-Speech technologies rather than depending entirely on general-purpose global APIs. - The system is designed to handle fast speech, pronunciation variation and proper-name errors using large language models and personalized voice-learning functions. - DABAR’s feature set includes Speech Cast for one-to-many interpretation, Link for one-on-one consultations, multilingual docent support, mobile receivers and multilingual meeting rooms. - SoyNet also built a hybrid interpretation platform for official events, where machine interpretation works with human interpreters and stenographers to correct errors in real time. - The company says it has used that hybrid model successfully at major events including GSAT in Changwon. - SoyNet has introduced an on-premise interpretation server that runs entirely within closed networks to keep corporate and government data from leaving the local environment. - The company says the on-premise system is drawing interest from tech firms, government ministries, universities with large international student populations and religious organizations. - Park said SoyNet started with AI inference accelerators that could speed up LLM training, cut latency and reduce memory requirements to one-fifteenth. - Park said the company’s early AI technology was about three times faster than Google’s inference speed. - The DABAR name comes from a Hebrew word meaning “to speak” or “word.” - The company says presenters can pre-upload materials so DABAR can learn terminology in advance before live events. - The service is already used in religious settings, including a church service in Australia where a QR code delivered Korean sermon translation to a phone. - Park said SoyNet expects about 1,000 churches to adopt the service in the near future. - South Korea has more than 2.71 million migrants, a figure SoyNet cites to support demand for multilingual services in religious and immigrant communities. - SoyNet sees the highest value opportunity in the MICE industry and points to markets such as Singapore, Dubai, Saudi Arabia, Thailand and Indonesia. - The company says its conference interpretation business is already producing consistent revenue. - SoyNet says its technology has been field-tested at events including the Global Startup Summit, G-START and GSAT 2026.

Between the lines: - The strategy combines product differentiation and market segmentation: high-security on-premise deployments for sensitive users, plus lower-cost hybrid tools for events and institutions. - By building its own core STT, MT and TTS stack, SoyNet is positioning DABAR as more controllable on pricing and data handling than API-dependent rivals. - The church and migrant-community pitch suggests SoyNet is looking beyond enterprise clients to everyday multilingual use cases where cost may be the main barrier. - Pangyo Techno Valley appears to be part of the company’s growth story, giving SoyNet access to talent, support programs and startup infrastructure.

What's next: - SoyNet plans to use new funding to scale server capacity and expand its sales team. - The company is pushing harder into overseas MICE markets and other multilingual hubs where live interpretation demand is high. - Growth will likely depend on whether SoyNet can convert event pilots and institutional interest into broader recurring deployments.

The bottom line: - SoyNet is betting that the next wave of interpretation tech will be private, customizable and built for live events, not just cloud convenience.