Gradium Launches stt-translate and s2s-translate, Real-Time Speech Translation Models Beating gpt-realtime-translate on Accuracy and Latency

Short summary

Gradium released real-time speech translation models (stt-translate, s2s-translate) supporting 5 languages across 20 pairs, claiming superior accuracy-latency tradeoffs versus OpenAI and Google alternatives. The architecture consolidates traditional three-model cascades into a single-pass system with voice selection and cloning features.

• •Gradium launched two real-time speech translation models covering English, French, German, Spanish, and Portuguese across 20 language pairs
• •Claims better accuracy-latency tradeoff than OpenAI's gpt-realtime-translate and Google's gemini-3.5-live-translate with simplified architecture
• •Includes output voice selection and cloning capabilities over duplex WebSocket

Generated with AI, which can make mistakes.