Google prides itself on having great machine learning capabilities, especially when it comes to decomposing and optimizing models large enough to run locally on your phone. The new Google Pixel 8 and 8 Pro are chock full of new features in this area. What I noticed right away was the addition of advanced local voice dictation for multilingual users. This is a feature that was previously limited to a single language on Pixel 6 and 7.
As someone who constantly switches between German and English when talking to my wife, I knew I had to get this and try it out for myself. The new features are impressive and better than anything we’ve seen on other phones, but there’s still a long way to go to achieve truly great multilingual voice input.
The path to multilingual voice input on devices
Before we get into the main topic, let’s briefly review voice input on Pixel smartphones. For the first time in his Google Pixel 6 and Android 12 in 2021, Google introduced a more advanced voice dictation branded as “Assistant Voice Input”. As with Google’s previous voice dictation method, Google servers do the heavy lifting and the audio is analyzed and transcribed locally on your phone. This is the same magic that turns audio into actionable text in Google Recorder, and it’s also what powers the auto-generated subtitles on your Pixel phone.
However, for a multilingual person like me, this solution is not ideal. When using the Gboard keyboard in multiple languages (like German and English in my case), voice dictation reverts to the old method and requires an active internet connection for best results. (There is a local fallback in case of airplane mode or poor connectivity, but it is more error-prone than other methods.)
The color of the microphone button tells you which audio input method is being used. When Assistant voice input is active, the distinctive four-color Assistant bar appears within the microphone icon and flashes as you speak. Otherwise, you’ll see a simpler, single-color animation.
Test your new multilingual assistant’s voice input
To get a general idea of how multilingual input on the Pixel 8 compares to others, we took a look at some of our bilingual chat messages to find samples that you can try your hand at voice input. However, some changes have been made for clarity. Then set up a Google Pixel 8 with the new multilingual assistant, a Google Pixel 7 Pro with the old multilingual typing, and an Apple iPhone 13 to see how good or bad things are on the other side of the fence. Did.
My method is by no means scientific, but I still wanted some comparability, so I used one voice prompt for all three devices and had their microphones on at the same time. That way, discrepancies between devices are not due to slightly different pronunciations or volumes across experiments. All three phones work with the same input.
Watch the video of the process above or read the transcription directly in the table below.
| no | original | pixel 8 | pixel 7 pro | iPhone13 |
| 1 | Hey, could you buy me some Brot und Käse at the supermarket today? | Hey, could you buy me some Brot und Käse at the supermarket today? | Hey, could you please buy me a broad-on case at the supermarket today? | Hey, could you pick up some Brodt und Käse from the supermarket today? |
| 2 | I also don’t know much about Gurken, Salat, and Tomaten. I think there must be some left, but they may be gone. | I am Gurken Tomaten Sarat, not you.I think there’s still a little left, but it might be gone. | I’m not you either, but I’m Gurken Tomatensarat.I think I should leave some of it, but I might not be able to do it anymore. | Also, Bad Gurken Tomaten Salat als You think you’re gone, but you’re not gone anymore |
| 3 | For dinner, we recommend focaccia, baguette, or pizza baguette. | Maybe something like dinner. | Maybe for dinner something like Katja oder Baguette gerne auch Pizza Baguette | Maybe for dinner something like Katja Oder Baguette Geln Ausch Pizza Baguette |
| Four | it’s okay. Bin Schon am Bahnhof has both works. | It’s okay, Bin Schon am Bahnhof’s work is amazing | All the good works of Bin Schön am Bahnhof Wurzburg | All is well, Bin Schon am Bahnhof Baus, works |
| Five | I think the next train will come soon, einer ist gerade weggefahren. | I think the next train will come within 2 seconds.Einar Ist Gelade Wegefalen | I think the next train will come soon.I noticed Vega Farm. | Show next train This comes second einer ist gerade weggefahren |
| 6 | Is this the way to get Persona auf den Gleizen? | Is this? How can I get Persona auf den Gleizen? | Is this how to get Persona auf den Gleizen | people who love greisen |
| 7 | I’m looking at Reshnung now, but I can’t see anything. | I’m looking at Rechnung now, but I can’t see anything. | I’m looking at Berechnung now, but I can’t see anything. | Ein look at Reshnung now and do nothing for every hundred. |
Looking at the bigger picture, it’s clear that the Pixel 8 wins this comparison, albeit by a small margin. No phone could get everything right, but the Pixel 8 was the one that most consistently recognized and respected language switches unless the keyboard switched randomly (more on this later) I will explain).
At the same time, the Pixel 7 Pro, which uses the older voice input method, comes surprisingly close, only failing most when it comes to ad-hoc language switching recognition. This is especially noticeable in examples 1 and 5, where the English is almost perfect, but the German part is gibberish in English.
Another thing I like about the Pixel 8’s voice input is that it adds punctuation, which older models don’t have yet. As for voice input, I would argue that this is a must. Because otherwise you’ll always have to go back and manually edit the transcript before sending it (or you just don’t care if others can understand it). .
There’s no need to mention the iPhone. The iPhone is horribly inferior when compared to both his Pixel smartphones, and is not recommended for anyone even trying to make their iPhone multilingual. I don’t understand how these errors occur, like the iPhone completely misspelling some words (‘lookat’ in 7) or making up other words (‘Eiting’ in 5) .
Switching keyboard layout becomes unstable
An even bigger problem with the Pixel 8’s new dictation method is that Gboard sometimes decides to arbitrarily switch between on-screen keyboards when using multilingual voice input. Particularly on sentence number 3, the Assistant’s voice input would consistently switch from the selected German/English keyboard to the dedicated English keyboard, completely ignoring my voice input following the first half of the sentence. I repeated this sentence many times and definitely encountered this behavior.
This erratic keyboard switching is also the reason for repeating one of the sentences in the Pixel 8 video. Because I didn’t immediately notice that my phone had switched to an English-only keyboard.
It seems like a new feature as I have never seen this kind of behavior on my phone before. After further testing, we were able to confirm that Gboard automatically switches the keyboard layout to the language that the user is primarily using at any given time. It may seem convenient at first glance, but I don’t think it is.
The big problem with this is that you can no longer force Gboard to listen to only one language. Even if only one keyboard is specially configured. without it For multilingual input, this setting is not taken into account by Assistant voice input. If you have a strong accent, you may run into problems where your assistant may misinterpret German-sounding English as German (as happened to me). Voice input is a big reason why I use a dedicated English-only keyboard that doesn’t understand German. Pixel 8 breaks this workflow.
This also affects mixed use of voice input and keyboard use. If the user selects a dedicated English keyboard to initiate voice dictation, the software must respect this explicit choice and not just switch randomly without interaction. After all, when you switch keyboards, the button layout also changes slightly. German QWERTZ layout is slightly different from American QWERTYit can be unpleasant when it happens without you realizing it.
It’s a long journey ahead, but we’re getting there
The unreliable full keyboard switching behavior is my biggest remaining issue with multilingual assistant voice input. More than that, it’s the best bilingual experience I’ve ever had with voice input. At the same time, look at the example above. The Pixel 8 may be the most accurate of them, but its transcriptions still have glaring mistakes and things that don’t make sense. But this is definitely a step in the right direction, especially considering the state of multilingual voice input just three years ago.
As generative large-scale language models mature, speech dictation could become an interesting use case. Once this kind of AI requires less energy and can work reliably and quickly on your phone, Gboard will simply perform voice input through his Google Bard model and fix any remaining errors. It may be. Gboard’s Proofread AI helps you correct spelling and grammar in your messages, and features like this are already in the works to help you retroactively correct your writing. For voice input, this step may need to be built directly into the workflow in the future to make this type of input truly seamless.
At the same time, Gboard is nearly multilingual actual It types for me and switches between autocorrect models for each language naturally and predictably. Gboard is probably the best Google app I use regularly, and it rarely disappoints me. Whenever I use an iPhone, I really miss the near-perfect auto-correction of my Android phone, even after iOS 17’s big (and necessary) upgrade in this department. We can only hope that Google continues to innovate and stay ahead of its competitors when it comes to inputs.
google pixel 8
If you want to experience the new multilingual assistant’s voice input, the Google Pixel 8 is currently your best bet. While this feature will later make its way to the Pixel 7 series, the Pixel 8 is still a significant upgrade over last year’s phones, with a better processor, perfect form factor, and sleek design. .
Source: Google
google pixel 8 pro
The Google Pixel 8 Pro is a more expensive route to take if you want Assistant voice input in multiple languages, but it also has more features than its smaller sibling, including a larger, brighter screen, more powerful camera settings, and additional software features. We are prepared.
