YomiPlay Is Back on the China App Store

YomiPlay is available again on the App Store in mainland China, and you can download it reliably now.

If you watched the demo, wanted to try it and couldn't find it, open the App Store and search for "YomiPlay" — it's there.

October 24, Programmers' Day. I had just polished YomiPlay's speech recognition to the point I was happiest with and was about to invite everyone to try it, when it was removed from the China App Store. At the end of that day's post I wrote: "After all, I really have put my heart into making it."

Being able to come back today with good news is a relief. This time I want to take it slowly and walk you through the details I never got to share last time.

To hear a sentence again, just tap that line

When I listen to Japanese podcasts on my commute, this happens all the time: I reach the end of a sentence and only then realize I missed a word in the middle. I drag the progress bar — too little and I miss the start of the sentence, too much and I have to sit through what came before. After going back and forth a few times, I sometimes forget which word I wanted to hear in the first place.

I remember thinking: if only I could just tap that sentence.

Now, import a podcast you like into YomiPlay and AI gives the audio text subtitles. Wherever you are in the audio, the subtitles follow. Want to hear a sentence again? Tap that line and the audio starts from the beginning of it. When someone talks fast, turn on single-sentence repeat and you can take your hands off the screen.

Some people like to hide the translation first. If they roughly understood, they keep listening. If they're not sure, they open the translation for a quick look. Sometimes a whole sentence hinges on one word — double-tap it and AI explains it in the context of the sentence. Once the meaning is clear, hide the translation again and shadow the original a few times, following its tone and pauses.

Some Japanese I only truly understood after the meeting

YomiPlay helps me at work too. Sometimes after a Teams meeting I keep thinking about something a colleague said. I roughly understood it during the meeting, but when I want to use that expression myself, I can't recall exactly how they put it.

When there's a meeting recording I can use for a review, I let YomiPlay turn what everyone said into subtitles and listen again. With the text in front of me, the sentence I've been thinking about is easy to find. One tap, and I hear how my colleague opened, how they made the question clear, how they naturally handed the conversation over to the next person. Some expressions I had seen in textbooks many times — only then did I know how I could use them in my next meeting.

Wherever your material comes from, you can import it from there. A screen recording in your photo library? Pick it from Photos. Audio or video in the Files app? Choose the file directly. A direct link to audio or video? Import it by URL. A ZIP of study material a friend sent you, or material shared in the YomiPlay community, imports with one tap — open it and pick up where you left off.

From on-device recognition to the cloud

When I started YomiPlay, my idea was simple: could the phone recognize speech by itself, and save the money of calling a cloud service? Later, the more I used it, the more specific my wishes became. For a full podcast episode, I wanted recognition to finish sooner. And for the sentence I couldn't catch in a meeting, I wanted the subtitles to be more accurate.

For those two things, I brought in large cloud models and more advanced AI services. Testing them over and over took a lot of time, and the servers and API calls are an ongoing cost. For me, seeing fewer things to fix by hand after generating subtitles for a recording makes it worth it.

If you want to try the new speech recognition, I recommend cloud recognition first.

Especially for a full podcast episode or a longer screen recording, the large model on the server is usually more accurate and faster. Most of the work happens on the server, which takes load off your phone's processor and battery. Of course, if you have an iPhone 15 Pro Max or a newer, similarly powerful model, on-device recognition works very well too.

Once you understand it, leave the subtitles blank for a few seconds

There's a way of practicing I'd really like you to try.

In Intensive Listening mode, have YomiPlay read the translation first, without looking at the Japanese. After the translation, give yourself a few seconds to think: if it were my turn to say this, how would I say it?

Try saying it yourself first, even if you stumble. Then listen to the Japanese original twice. The first time, listen for whether they used the words you thought of. The second time, notice where they pause, and how they get through in one breath the part where you just stumbled. Then say it yourself once more.

If a while later you can't remember it again, put the sentence in your deck. When spaced repetition brings it back for review, give yourself another chance to say it.

I think the happiest moment of learning a sentence often comes a little later. Maybe a few days on, someone in a meeting happens to ask you a question, and the Japanese you practiced just comes out.

That's the moment it truly becomes your own words.

Even if all you have is text, you can listen to it

Sometimes I read a Japanese piece I like. I get the gist, but there are a few words I don't know how to pronounce. Paste the text into "Text to Speech" on the import page and you get natural narration, with floating subtitles synced to the voice.

The subtitles you save work just like the ones from an imported podcast. To hear the sentence a word is in, tap that line; you can repeat, listen intensively and shadow too. An article you read at night goes with you when you head out the next day.

Because subtitles get read over and over, I left editing in the hands of the people using them. If a compound word is split wrongly, you can fix the segmentation by hand. If a kanji should be read differently in this sentence, you can change that yourself. For kanji with several readings that you run into often, you can save your usual reading in "My furigana dictionary", which makes the next fix easier.

If you're learning Japanese, you may know this feeling: you stare at a long string of katakana and still can't place it, and the moment you see the original English word, it clicks. So YomiPlay also shows the English origin and meaning of katakana loanwords. When tidying up subtitles, you can have SI help with proofreading and translation, then go through each spot with your own understanding. Ha — note that this SI (super intelligence) is one I genuinely pay for. You can even pick a subtitle color theme that's easy on your eyes.

Words worth keeping sometimes find you along the way

When you see a Japanese sentence you want to remember, snap a photo and put it in your deck. When you come across it again some days later, you may still remember where you saw it. A deck you've put together can be shared with friends learning Japanese too.

Podcasts you like can be shared as well. Send the whole set of subtitles to a friend to study together, or pick out just the best short clip, export it as a subtitled video and post it on a video platform. Even a fun word or a sentence that made you smile from your deck can become its own video, for that person who'll get it at a glance.

When YomiPlay was removed from the China App Store last time, the hardest part was that I was just about to say "you can download it and try it", and couldn't. Today I finally can. And thank you to everyone who sent me screenshots and carefully described problems. Many of the small annoyances you pointed out have since become the parts of YomiPlay that feel smoother to use.

On your next commute, bring along a podcast you already wanted to listen to. When you hit a sentence you like, stop there for a moment. Hear it clearly, then say it yourself.

If one trip lets you take away, or leave behind, a single sentence, that's plenty.

Oh, and one more thing — an Easter egg

I've written all this about Japanese, but you can also put an English podcast you like into YomiPlay, pick an English sentence you want to learn, tap it to listen, and say it after the speaker. And if you have a Japanese friend who's learning Chinese, ask them to try it too. Usually you're the one learning Japanese from them; this time, it's their turn to practice with a Chinese sentence you shared.

If you find this Easter egg, come back and tell me which language you learned with YomiPlay, and which sentence you remembered. You're welcome to share that sentence with everyone, too. Maybe the next person is hoping to learn it.