Hands-on with Apple Intelligence in China: Writing Tools, Visual Intelligence, and Live Translation Tested

November 4, 2026 · 16 min read

Hands-on with Apple Intelligence in China: These features are ready

Lead author: Wantong, a Minorit author and a famous wanderer of the Greater Bay Area who needs to sleep more than ten hours a day

WeChat official account: Anna's Five Hundred Pounds Outlying Island

Wantong, Wantong, a Minorit author and a famous wanderer of the Greater Bay Area who needs to sleep more than ten hours a day

WeChat official account: Anna's Five Hundred Pounds Outlying Island

Co-author: Wantong, a Minorit author and a famous wanderer of the Greater Bay Area who needs to sleep more than ten hours a day

WeChat official account: Anna's Five Hundred Pounds Outlying Island

Wantong, Wantong, a Minorit author and a famous wanderer of the Greater Bay Area who needs to sleep more than ten hours a day

WeChat official account: Anna's Five Hundred Pounds Outlying Island

On September 10, 2024, with the release of the iPhone 16 series, Apple's China website put up the slogan "Ready for Apple Intelligence." Who could have guessed that this "preparation" would last a full 785 days. Now, the long-delayed Apple Intelligence for China has finally arrived today.

According to the schedule Apple announced, the first batch of core features will begin testing today in the latest iOS 27.3 developer preview, and is planned to launch with the official iOS 27.3 release in December. The first batch of features will be enabled on iPhone first, including Writing Tools, Clean Up in Photos, and Live Translation. At the same time, Siri will also officially connect to Qwen and Baidu, bringing stronger answering capabilities.

However, what has been approved this time is still the initial version of Apple Intelligence with the "marquee" Siri interface. The Dynamic Island liquid glass interface unveiled at WWDC 2026, as well as the next generation of Apple Intelligence and Siri AI features, are not included in this test.

But regardless, the long wait has finally produced a result. So we also updated to the iOS 27.3 beta at the first opportunity to see how they actually work.

Which devices can use it?

This Apple Intelligence beta is only available on iPhone, and supported models include iPhone 15 Pro, iPhone 15 Pro Max, and all iPhones released from iPhone 16 onward, with iPhone 16e and iPhone 17e also included. As for iPhone 15 and iPhone 15 Plus, because they only have 6 GB of memory, they are still left out this time. Other devices such as iPad and Mac are not yet supported and still need to wait for the corresponding approval.

For an iPhone that meets the model requirements, to enable Apple Intelligence for China, you only need to meet any one of the following conditions:

Use a China-market device, that is, a device where the last three characters of "Model" in "Settings > General > About" are "CH/A";

Use a device from overseas, with the Apple Account signed in set to country or region China mainland, commonly known as a "China region ID";

After meeting the above conditions and updating to the iOS 27.3 developer preview, the former "Siri" in Settings will become "Apple Intelligence & Siri." After entering, tap "Turn on Apple Intelligence" at the bottom, and you will see the button to get Apple Intelligence.

According to Apple's introduction, up to 7GB of Apple Intelligence support content may need to be downloaded, and after the download is complete, you also need to wait for Siri to upgrade and be enabled, with the required time varying. Fortunately, there is no waitlist this time, so you do not need to queue first for access.

Taking my iPhone Air as an example, before the update, iOS and system data occupied 17.01 GB and 27.14 GB respectively. After updating to the iOS 27.3 beta and enabling Apple Intelligence, they became 22.69 GB and 31.59 GB respectively, of which Apple Intelligence occupied 5.92 GB.

Siri marquee and third-party model integration

The most intuitive change in this Apple Intelligence is undoubtedly the "marquee" effect when summoning Siri. However, in addition to long-pressing the side button to interact with Siri by voice, you can also double-tap the bottom navigation bar on the screen to type directly and talk to it.

In terms of models, Apple Intelligence for China only has an on-device model, used for basic Siri functions, Clean Up in Photos, Live Translation, and other tasks. As for more complex requests that the on-device model cannot handle, the question and context will be sent to Alibaba's Qwen model for processing.

However, the Qwen extension is not enabled by default and needs to be turned on manually. You can go to "Settings > Apple Intelligence & Siri," tap "Qwen" under "Extensions" below, then tap "Use Qwen," and follow the prompts on the pop-up page to complete account login to enable it. In addition to logging in to an existing Qwen account with a phone number, you can also choose "Sign in with Apple" to hide your email.

After enabling it, with Apple Intelligence's ability to recognize screen content, you can directly ask Qwen through Siri about photos, PDFs, or presentations displayed on the screen, and continue the conversation.

You can also view the history of conversations with Qwen through Siri in the Qwen app, provided you are signed in to the same account. In addition, Qwen will also appear in the Visual Intelligence and writing tools introduced later, where it is used to understand images, answer questions, and create text and images.

As for when data is sent to Qwen, Apple says user consent is required before each request is sent. In my tests, whenever Siri needed to call on Qwen for an answer, it did indeed first ask whether to allow it, and only sent data after receiving consent. Apple also says that data sent to Qwen through Siri will not be used for model training, marketing advertising, or user tracking.

Besides Qwen, Apple Intelligence in China has also partnered with Baidu, though currently this is limited to image search and is not responsible for Siri's question answering.

In short, with the addition of large models, Siri is finally no longer the naive robot that used to say "Here's what I found on the web" at the drop of a hat, and its language understanding has improved considerably. It can better understand pauses and corrections mid-speech, and can handle follow-up questions in context. For example, if you realize mid-sentence that you said the time wrong, you can correct it directly without having to cancel the conversation and start over as before.

But why did it end up in the work to-do list (

Writing Tools

In daily life, we inevitably deal with text. Writing reports for the boss, replying to different people, or not wanting to read a long document from the beginning—all of these can be handled through the writing tools.

As long as an app uses the iOS system's built-in text selection menu, including third-party apps and websites, you can select text and tap the Apple Intelligence icon in the pop-up menu to use the writing tools. However, apps like "Xiao Er Mei" that have rewritten the menu themselves cannot use it for now.

For text content that cannot be selected, you can also use screenshot OCR to select text on the screen and then invoke it.

English text is being prepared.

After making changes, you can also see what was changed and why. Of course, AI uses more standard and formal grammar, so for some playful, meme-style writing, it may also think you made a mistake. Therefore, if you are not satisfied with the AI's rewrite, you can keep the original wording.

If you want to express the same meaning in a different way, you can use the "Rewrite" feature. How should I put it? This feature is very much like the key question type in my middle school English exams, "sentence explanation," that is, synonym sentence conversion. (If you were also a middle school student in Shenzhen around 2010, you should remember it vividly.) When rewriting, Apple Intelligence understands the meaning of the original sentence and then writes it out in a different way.

Of course, due to limited capability, the on-device model may produce some abstract stuff...

In addition, you can choose among three tones—"Friendly," "Formal," and "Concise"—to adjust wording according to the occasion. For example, the message you need to convey to Party A and Party B may be the same, but the wording may require more careful consideration—copy sent to Party A can be sent to Party B, but copy sent to Party B may not necessarily be sent directly to Party A.

If none of the above rewrites meet your needs, or if you need to write a new piece of content from scratch, you can also use the "Compose" feature. Swipe up on the writing tools panel and tap "Compose" to enter free creation mode. Enter your creation requirements in the input field, and Apple Intelligence will call on Qwen to create according to your requirements.

In addition, the writing tools also provide a summary feature, condensing long content into four forms: summary, key points, list, and table. When you encounter something too verbose (such as the big promises your boss makes in meetings), you can have it organize it first to grasp the key points at a glance, or discover that there are no key points.

Visual Intelligence

Visual Intelligence is also one of the core features of Apple Intelligence this time. For iPhones equipped with a Camera Control button, you can start Visual Intelligence by long-pressing that button; if you are using an iPhone 15 Pro series without a Camera Control button, you need to first add the "Visual Intelligence" button in Control Center and then enter through Control Center.

After starting Visual Intelligence, point the lens at the object you want to recognize, press the Camera Control button or the shutter button on the screen to freeze the current frame, and choose "Ask" or "Search."

Tapping "Ask" on the left sends the image to Qwen, which generates an answer based on the image content and your question. In addition, you can continue asking follow-up questions after receiving an answer. In my view, this feature is especially useful when traveling abroad. For example, if you see an introduction to an exhibit you like in a museum, you can simply point the lens at it to recognize the text and further ask about the specific content.

"Search" on the right sends the frozen frame to Baidu Image Search to find related results based on the image content. If there are multiple objects in the frame, you can also circle the one you want to search for.

In addition to recognizing objects, Visual Intelligence can also directly process text in the frame. For example, when a large block of text appears in the frame, Visual Intelligence can recognize the content and directly generate a summary, saving the steps of manual copying and pasting.

Image Processing

I believe many people have had this experience: tens of thousands of photos are stored on their phone, but when they want to find a particular one, they simply can't. They clearly remember the scene in the photo, but can't recall exactly when it was taken, so they can only flip through the gallery one by one.

With the arrival of Apple Intelligence this time, we can finally use natural language search in the Photos app on iOS.

However, in actual use, the accuracy of this feature has not yet met my expectations. For example, when I directly search for "photos of sunrise at the beach," the system can indeed find the corresponding images based on the description, but when faced with more complex descriptions, the search results are not so ideal. Sometimes unrelated photos appear, and sometimes photos that clearly meet the criteria do not appear in the results.

Considering that my photo library is still in the analysis stage, perhaps I need to wait for the system to finish analyzing before I can further judge the actual performance of this feature.

In addition to searching photos, Apple Intelligence also gives "Memories" in the Photos app customization capabilities. In the past, memory videos were mainly generated automatically by the system, and the content users could decide was relatively limited. Now, you only need to enter a natural language description, and the system will select eligible photos and videos from the library based on the theme, and arrange these materials into a memory video with a sense of narrative.

Beyond these features, what I look forward to most is "Clean Up" in Photos.

This feature is driven by an on-device model and can be used even without a network connection. After enabling Apple Intelligence, enter the photo editing interface, wait for the download to complete, and you will see the "Clean Up" option. Just tap, brush, or circle to remove objects in the frame that you do not want to keep.

However, expectations aside, the actual results are a matter of opinion.

In my experience, if the photo background is relatively clean, the image information is relatively simple, or the area to be removed is small, the result after removal is usually fairly natural, and it is hard to notice obvious processing traces without looking carefully.

But if the background contains a large amount of complex texture, or the object to be removed blocks a large area of the image, the filled-in content easily gives itself away, and obvious errors may even appear.

Open any conversation in the Messages app and tap the "+" sign on the left to find the entry for Genmoji. You can choose from the suggestions provided by the system, or directly describe the emoji you want in natural language. You can even use friends and family in your photo library as prototypes, pairing them with different hairstyles, expressions, and styles to add some personalized expression to everyday chats.

Live Translation

Live Translation may be one of the most important features in this Apple Intelligence update. Like "Clean Up" in Photos, it is also driven by Apple's on-device model, and after downloading the required models and languages in advance, it can be used even without a network connection.

Although the device and Siri language support range of this Apple Intelligence update includes Simplified Chinese and English, Live Translation supports a richer range of languages. In addition to Chinese and English, it also covers French, Japanese, German, Spanish, and many other languages. Live Translation supports Messages, Phone, FaceTime, and face-to-face communication through AirPods.

In the Messages app, users can directly translate conversation content; during FaceTime video calls, the translation results appear on the screen as subtitles; in ordinary phone calls, the system plays the translated content through voice, allowing both parties who speak different languages to communicate smoothly. For face-to-face communication, Live Translation can also directly translate what the other person says into your own language and play it through AirPods, so you no longer need to check the translation on your phone.

Currently, face-to-face AirPods Live Translation supports AirPods 5, AirPods 4 (Active Noise Cancellation), AirPods Pro 2 and later models, and AirPods Max 2 with the latest firmware. If you are a user of these AirPods, once they are paired with an iPhone that has Apple Intelligence enabled, you will see the "Live" page entry in the Translate app.

As for the actual effect, in my experience, Live Translation can already correctly understand and translate most conversation content, and basic information transmission is not a problem. But the translated text sometimes still seems rather stiff, especially in some more colloquial expressions, where the traces of machine translation can be clearly felt.

Although there is still a certain gap before natural and fluent cross-language communication, for everyday communication it already has quite good practical value.

Summaries and Attention

Compared with Live Translation and photo Clean Up, the changes Apple Intelligence brings to everyday features such as notifications and mail may not be so noticeable, but they are equally practical. They can reduce the time we spend processing information and make it less easy for our attention to be interrupted, which is indeed very friendly to someone like me with AuDHD.

In terms of notifications, Apple Intelligence analyzes all notifications received by the iPhone and displays the content that needs priority attention at the top of the notification list in the form of "Priority Notifications." At the same time, for multiple notifications sent by the same app, the system can also automatically extract the key points and generate a short summary, so there is no need to check them one by one.

However, the "Generate Notification Summaries" feature is not enabled by default. You need to go to Settings > Notifications > Generate Notification Summaries to manually enable and configure it.

Meanwhile, the "Reduce Interruptions" Focus mode has also officially launched with the arrival of Apple Intelligence. Unlike traditional Focus modes, which mainly rely on preset rules to filter notifications, the "Reduce Interruptions" mode backed by Apple Intelligence can determine whether the content of a notification is important, prioritize information that may need to be handled promptly, and temporarily silence less urgent notifications, minimizing unnecessary interruptions.

The Mail app has also seen similar changes. When faced with a long email or a lengthy back-and-forth email thread, Apple Intelligence can directly extract the main content, helping us quickly grasp the key points. "Priority Mail," meanwhile, places time-sensitive emails that need prompt attention at the top of the inbox, preventing important information from being buried in a flood of messages.

In addition, recordings in Notes and call transcriptions in the Phone app also support generating content summaries through Apple Intelligence, saving you the trouble of listening to an entire recording again or reading through large blocks of text.

Final Thoughts

It is not hard to see that the arrival of Apple Intelligence this time has not brought many surprises. For users in China, the long wait has already worn away their expectations. During the wait, the generally negative reviews of the initial version of Apple Intelligence from overseas users also gave us psychological expectations for its actual performance.

After all, when AI from various companies is evolving toward Agents, and my own workflow has already been greatly accelerated by the addition of Agents, going back to experience these features focused on information summarization, image processing, and simple interactions inevitably gives a feeling that it is too little, too late.

I can only say that the arrival of Apple Intelligence for the China market this time mainly fills the gap of Apple devices "having no AI." It still has a long way to go before it is sufficiently useful and practical, or even as eye-catching and "good-looking" as the new Apple Intelligence liquid glass Siri UI.

Of course, if you still cannot resist trying it out, a reminder is in order: what is currently being pushed is still a beta version, and trying it out carries risks, so remember to back up before updating.

> Follow sspai on Xiaohongshu to experience a wonderful digital life 🍃

> Practical, useful genuine software, presented to you by sspai 🚀Apple Artificial Intelligence iPhone This article's editor:

@Guangling Zhixi

8 fellow readers have already charged