Apple Intelligence Adds Real-Time Translation to FaceTime and iMessage in iOS 20

Apple's approach to AI has always been different from everyone else's. While the rest of the industry has been racing to build ever-larger language models and deploy them in chat interfaces, Apple has been quietly baking AI into the places where people already spend their time. iOS 20, announced at WWDC 2026 and rolling out to users now, contains what is probably the most practically useful Apple Intelligence feature to date: real-time translation built directly into FaceTime and iMessage, with no third-party app, no copying and pasting, and no internet connection required for 12 supported languages.

The feature works differently depending on the context. In FaceTime, it operates as a live caption overlay that shows both sides of a conversation translated in near real time, with a reported latency of under two seconds for the 12 on-device languages. iMessage gets an automatic translation toggle that can display incoming messages in your preferred language while keeping a tap-to-reveal option showing the original text. There is also a compose mode that lets you write in your own language and send in the recipient's language, with a preview of the translated text before you hit send.

The on-device capability is the genuinely impressive part. Apple has been building out its on-device neural engine specifically to enable features like this, and the translation quality for the on-device languages, which include English, Mandarin, French, Spanish, German, Japanese, Korean, Arabic, Portuguese, Italian, Dutch, and Hindi, is competitive with cloud-based services for most everyday conversational content. Apple hasn't published BLEU scores or formal benchmarks, which is consistent with the company's usual reluctance to make its AI claims directly quantitatively comparable. But early testing by language researchers has been largely positive.

Why On-Device Matters More Than People Realise

Cloud-based translation is fine when you have a reliable connection and are happy with your conversation data leaving your device. In a FaceTime call between, say, a patient and a healthcare provider who speak different languages, or between a lawyer and a client, or in any number of sensitive professional contexts, cloud-based translation raises real data concerns. Apple's on-device approach sidesteps those concerns entirely, which opens up use cases that cloud alternatives simply cannot address.

This is the strategic logic underlying Apple's entire approach to Apple Intelligence. By keeping as much as possible on-device, Apple can offer features in contexts where competitors with cloud-first architectures face genuine barriers. Healthcare, legal, education, and enterprise environments all have data sensitivity requirements that on-device processing handles neatly.

The tradeoff is capability ceiling. The on-device models Apple is running are genuinely impressive given the hardware constraints, but they are not going to match a frontier model running in a data centre on unlimited compute. For translation of complex idiomatic language, technical documents, or low-resource languages not in the supported 12, the on-device results will be noticeably worse. Apple routes those cases to its Private Compute Cloud, which uses server-side models with privacy protections the company has invested heavily in auditing and publicising.

The Broader iOS 20 Intelligence Picture

Real-time translation is the headline Apple Intelligence feature in iOS 20, but it ships alongside several others worth noting. The updated Siri can now handle multi-step tasks across apps with a success rate Apple claims exceeds 90 percent on its internal benchmark set, up from around 65 percent in iOS 19. Mail's priority inbox now uses on-device understanding of your communication patterns to surface messages, rather than simple heuristics. And Photos has added a semantic scene description feature that generates alt text for images being shared, which has immediate accessibility implications.

The cumulative picture of Apple Intelligence in iOS 20 is of a company that is becoming noticeably better at the slow, patient work of making AI useful in the specific grain of how people actually use their phones, rather than trying to slot a chat interface into everything.

What This Means for the Translation Market

Google Translate, Microsoft Translator, and dedicated apps like iTranslate have not lost their relevance overnight. For the 40-plus languages those apps cover that Apple's on-device feature doesn't, or for use on Android, they remain essential. But Apple quietly removing the need for a separate translation app for a significant percentage of the world's most common language pairs is a meaningful shift. This is how Apple tends to eat markets: not with a competitive product launch but by making a capability standard and invisible.