August 2026 newsletter

News, jobs, community questions and more about machine translation from August 2026

🗞️ News

The US Congress is debating the Language Access for All Act, which would restore language-access requirements revoked in 2025 and stop AI from replacing human translators.
Gartner features ModelFront in its 2026 Hype Cycle for Natural Language and Unstructured Content Technologies.

🚀 Launches

Google DeepMind launches SL2T, a sign-language-to-text model that adds ASL dictation to Pixel 11’s Gboard and Live Transcribe.
OpenAI launches GPT-Live, a voice model that replaces Advanced Voice Mode, listens and talks at the same time, and demoed live translation.
Eurovision Sport and CAMB.AI launch AI-translated live commentary at the European Athletics U18 Championships.
Anthropic launches multilingual support for Claude’s voice mode on Opus and Sonnet models.
Translated launches Lara 3, adding training based on human feedback plus image, audio, and document translation.
Instagram and Facebook launch Reels translation in 18 languages, with auto-dubbing and AI lip-sync.
TAUS launches version 3 of its EPIC quality-estimation model.
Threads launches translation support for Live Chats.
Amazon launches Kindle Translate, an AI translation service, in French, Italian, and Portuguese for self-published authors.
DeepL Voice launches real-time translated captions for Google Meet.
Pangeanic launches MTQE v2, a quality-estimation model, on Hugging Face.
Azure Translator API launches LLM choice, adaptive MT, and tone controls.

🛠️ Community launches

Wenyi, an open-source pipeline for consistent terminology and character names in Chinese novel translation.
GenSRT, a free, open-source Whisper subtitle app for Windows, with translation and burn-in.
Rhetoriq launches an app for AI rewriting, translation, and dialect conversion, adding High Valyrian for Game of Thrones fans.
XliffAI, an open-source CLI for translating XLIFF 2.x files with LLMs.
GianoReader adds side-by-side EPUB translation on phones and tablets.
Olaverse Lab launches DiacNet-1.1, a specialized model for restoring diacritics in 10 languages, fine-tuned from Google’s ByT5.
Libyx, a free e-book translator supporting 43 languages, cuts costs with batch mode and prompt caching, and refines quality with a two-model pass.
FTA, a browser extension that shows an LLM-powered translation in a popup over selected text, using Google AI Studio, OpenRouter, or local models.
Meeting Interpreter, a real-time speech translation app for Windows.
DocTranslating, a document translator that lets you pick between DeepL, Google, Azure, and Gemini per job, with OCR support for handwritten text.
OLVA, a bot-free AI meeting assistant, adds live translation in 87 languages across Zoom, Google Meet, and Microsoft Teams.

📚 Research

Taia Translations shares its experience adding LLM tiers to its product, citing unreliable segment handling and a shift to credit-based pricing.
WMT26’s quality-evaluation shared task closes for submissions, covering error detection, score prediction, and error-free classification.
A technical walkthrough of building a Transformer from scratch in PyTorch, for English-to-Tamil translation.

❔ Questions

How do you preserve table format when translating English to Spanish files?
Is a Google Lens translation trick real?
Which real-time voice translation app works for English and Hungarian?
Has DeepL gotten worse?
What AI instructions do people use for translation?
Can anyone beta-test a Hindi translation on a Lord of the Rings-themed site?
Can anyone help with a Sanskrit translation for a research project?
What advice do people have on preprocessing for an MT internship project?
Which local STT/ASR subtitle tool should I use?
What workflow do people use for translating long-form books with AI?
What do you learn building an LLM-based translation setup on a budget?
Which MT tools can translate medieval Tibetan into English?
Does machinetranslation.com actually work?
Why doesn’t WhatsApp use a voice transcriber as accurate as Instagram’s?