Latest updates · 21 October 2026
In the full issue →Speech AI is widening its language reach: India's Parliament will add Manipuri to its Sansad Bhashini translation system, and Amazon Polly's generative engine now speaks Hindi and Finnish.
Research, industry & policy
- Manipuri to be included in Sansad Bhashini, asiaPOST · 18 Oct 2026: Manipuri will be added to the AI translation platform for real-time translation of parliamentary proceedings.
- Amazon Polly Generative engine adds 14 voices and two languages, AWS · 20 Oct 2026: A new model brings more natural, low-latency speech, 14 new voices across nine locales and Hindi and Finnish on the Generative engine for the first time, including Hindi and Indian English voices Kiara and Arjun.
Journals & preprints
- DeepSEA: A Multilingual Evaluation Benchmark for Southeast Asia, arXiv 2610.26155 · 19 Oct 2026: While language models have demonstrated remarkable capabilities, their performance remains disproportionately skewed towards high-resource languages.
- Monsoon: Scalable Spontaneous Speech Collection for Low-Resource ASR in 50 Languages, arXiv 2610.26776 · 19 Oct 2026: For most languages, automatic speech recognition (ASR) is limited less by model architecture than by a lack of accurately transcribed spontaneous speech.
Why it matters
Adding languages to public and commercial speech systems is how AI reaches people outside the English-speaking web.
