OneVoice AI Challenge 2026
About OneVoice AI Challenge
Hosted by Saigon AI Hub × Qualcomm, OneVoice AI Challenge is a global Edge AI competition inviting researchers, engineers, startups, and technology builders to develop the next generation of real-time multilingual translation solutions powered entirely on-device.
As globalization accelerates across workplaces and communities, multilingual communication has become increasingly essential. Yet in many real-world environments such as factories, construction sites, logistics hubs, and service industries, language barriers persist, especially in situations with unstable Internet connectivity, hands-busy workflows, or limited accessibility to technology.
To address this challenge, OneVoice AI Challenge calls on innovators worldwide to design practical AI-powered translation devices that enable more seamless, intuitive, and accessible communication in everyday life.
Challenge Topic
Participants are challenged to research, design, and prototype standalone multilingual translation solutions powered by Edge AI, capable of processing directly on-device without relying on cloud infrastructure.
The Goal
Enable natural bilingual conversations with minimal latency, especially in multilingual communication environments with limited connectivity and real-world operational constraints.
Scoring criteria:
The competition evaluates entries based on three main pillars, with a heavy emphasis on technical execution:
Innovation & Novelty (25%): Assessment of your unique approach to problem-solving and any additional value-added features your solution provides.
Technical Performance (50%): Evaluation of the on-device translation system's accuracy, duration, speed, and latency.
Accuracy:
Automated Metrics: Systems will be evaluated against a proprietary benchmark dataset. Scores will be derived from BLEU (n-gram overlap) and COMET (neural-based framework).
Naturalness: A Mean Opinion Score (MOS) will be assigned based on a panel of native speakers evaluating the prosody and clarity of the synthesized speech.
Efficiency & Latency:
Real-time Factor (RTF): The system must maintain an RTF < 1.0.
Total Turnaround Latency: Measured as the duration from "End of Speech" to "Start of Audio Synthesis." Solutions with latency exceeding 2.0 seconds will receive significant point deductions.
On-Device Constraints: Models must run locally. Internet dependency during the testing phase will result in disqualification.
Business Viability (25%): Assessment of the product’s market readiness and its hardware sustainability.
Target Language Pairs
Participants may develop solutions for one or more of the following language pairs:
Vietnamese ↔ English
Vietnamese ↔ Mandarin Chinese
Vietnamese ↔ Korean
Device Format
Solutions may be implemented on any portable, handheld, wearable, or embedded device format. No predefined hardware category restrictions apply.
Competition Format
Team-based competition (no limitation on team size)
Global hybrid format:
Online development phases
Offline Grand Finale at VNG Campus, Ho Chi Minh City, Vietnam
Competition Timeline
24 May - 24 June 2026: Registration period
June - July 2026: Technical specification submission
September 2026: Prototype submission
October 2026: Field testing & evaluation
November 2026: Grand finale at VNG Campus
Who Should Join?
OneVoice AI Challenge welcomes innovators from across the global AI ecosystem, especially participants from Vietnam and Asia:
Independent researchers: AI practitioners, engineers, and applied researchers
Startups & small teams: Early-stage teams building practical AI solutions
University teams: Student and research groups combining academic research with applied innovation
🚀 Ready to build the future of multilingual communication with Edge AI?
Join OneVoice AI Challenge 2026 and turn your ideas into real-world AI solutions.
🔗 Learn more: saigonaihub.com/OneVoiceAIChallenge
🔗 Template for Participant/Team information: Here