Gradium Launches stt-translate and s2s-translate: Real-Time Speech Translation Models Outperform GPT and Gemini on Accuracy and Latency

Gradium has officially launched two new real-time speech translation models—`stt-translate` and `s2s-translate`—designed to deliver live, streaming translations directly in the browser. Both models support five languages and are engineered to provide faster and more accurate translations than existing solutions from major competitors. According to Gradium, their models achieve a superior accuracy-latency tradeoff compared to `gpt-realtime-translate` and `gemini-3.5-live-translate`. In addition, Gradium’s offerings include advanced output voice control—including voice cloning—a feature notably absent from `gpt-realtime-translate`. ## TL;DR - Gradium releases two real-time speech translation models: `stt-translate` and `s2s-translate`. - Both models stream results live in the browser across five languages. - Claims better accuracy and lower latency than `gpt-realtime-translate` and `gemini-3.5-live-translate`. - Adds output voice control, including cloning, a feature not available in `gpt-realtime-translate`.

via MarkTechPost