A dokáže AI překládat mluvenou řeč v reálném čase mezi hlavními jazyky ?

Appleovy překladatelské sluchátka, Google Pixel Buds Pro 2, Meta Ray-Ban — překlad řeči do řeči se v roce 2024 stal spotřebitelskou funkcí.

Background

Apple's translation earbuds, Google's Pixel Buds Pro 2, and Meta's Ray-Ban smart glasses have integrated speech-to-speech translation as a consumer feature as of 2024, making real-time interpretation accessible through wearable tech.

Current AI systems can translate spoken speech in real time across major languages by combining automatic speech recognition (ASR), machine translation (MT), and text-to-speech (TTS) synthesis. These systems process the spoken input, convert it to text, translate the text into the target language, and then synthesize the translated text back into speech, all within seconds. Recent advancements—particularly the development of end-to-end speech translation systems—have streamlined this pipeline, improving both speed and naturalness of the output.

While accuracy and fluency vary by language pair and context, research indicates steady progress in reducing errors and enhancing contextual understanding. Notable contributions to this field have come from both industry and academia, with frameworks like Whisper (for ASR) and models such as M2M-100 and NLLB (for MT) playing foundational roles. Benchmark evaluations continue to push the boundaries of real-time translation quality, especially for lower-resource languages.

Over the past five years, the combination of large-scale neural models and improved hardware has enabled near-instantaneous translation in everyday settings, from travel to professional communication. Ongoing work focuses on handling dialects, background noise, and emotional tone to further humanize the experience.

[IEEE, Enriched May 9, 2026]