Breaking Update
Deepdub Launches Ultra-Low Latency Voice Model for Enterprise Applications
Deepdub launches Lightning 2.5, ultra-low latency voice model for enterprise applications. Half-second faster latency, 2.8× throughput, 5× concurrency.
Jerusalem, 11 September, 2025 (TPS-IL) — Tel Aviv-based AI voice technology company Deepdub has launched Lightning 2.5, its fastest and most scalable real-time voice model to date.
Designed for enterprise applications, multilingual AI agents, and live content localization, Lightning 2.5 delivers latency as low as 200 milliseconds — half a second faster than industry standards—alongside 2.8× more throughput and 5× higher concurrency than its predecessor.
Built on Deepdub’s foundational voice AI models and optimized for NVIDIA GPUs, the system preserves speaker identity, intonation, and emotional nuance for lifelike, production-quality experiences. Deepdub will demonstrate the model at the upcoming International Broadcasting Convention 2025 in Amsterdam.
