Nothing, the consumer electronics startup known for its transparent-back phones, just shipped an on-device AI dictation tool supporting over 100 languages. Unlike most voice-to-text tools that send your audio to a remote server for processing, this one works entirely on the device itself.
That distinction matters in practice. Cloud-based dictation requires a working internet connection and means your audio gets processed on someone else's servers. On-device processing is faster for short bursts of text, works offline, and keeps recordings private. The tradeoff is that on-device AI models are typically smaller than their cloud equivalents, which can affect accuracy on accented speech or technical vocabulary.
The 100+ language count is the most interesting part of this announcement. Apple's Dictation supports around 60 languages, and many Android OEM voice tools cap out even lower. Covering 100+ languages suggests Nothing is using a multilingual model architecture - a single model trained across many languages at once - rather than bundling separate per-language models, which would be prohibitively large to fit on a phone.
Nothing has been steadily adding AI features to Nothing OS, competing for relevance against Samsung's Galaxy AI and Apple Intelligence. A dictation tool is one of those features that sounds routine but gets used constantly - voice notes, composing messages, accessibility workflows. The feature appears to be bundled into Nothing OS rather than sold separately, which is the right call for a brand still building its user base.
TechCrunch covered the announcement but Nothing hasn't published accuracy benchmarks comparing it to cloud-based alternatives. That's the number that would actually tell you whether to use it over your current setup.