A little feedback on 4.0

Alfred

Member
After trying out Pleco 4.0 beta for a few days last year, and a few more days recently, and comparing it with 3.2/Legacy, here is my feedback. For now I'm going to stay using 3.2 or Legacy. Here are the reasons why, based on my own particular way of using the app:


Audio

The main reason why I installed 4.0 beta is because I thought the audio would be improved. But it hasn't improved and in some aspects it's gotten worse.

The audio for each individual word is still very good, because I think it was recorded by a real human male and female speaker, so it's the same quality in 3.2 and 4.0.

But in real life, you never hear/say words individually. So the most important thing in audio is being able to hear the example sentences. Not only it allows you to practice real pronunciation, articulation and fluency as you need it in real life, but it also helps a lot in learning and remembering the word and its usage, as you hear it in the context of real life sentences.

In fact, reading the definitions of a word is something you do once or twice, when you first encounter a new word. But reviewing a word is something you have to do many more times, perhaps 10 or 20 times overtime, until you properly learn it, when it's part of your active vocabulary.

And when I review a word, I almost don't need to check out the definitions again, or I do it very quickly. Mainly, I go straight into the example sentences to practice it in context as you would hear it or say it in real life. That way you can visualize the meaning and keep getting more and more familiar with it, and with all the ways that word can be used.

So this is why the audio of the example sentences is far more important, in my opinion.

And in 4.0, the purchased versions of Hong and Qiang TTS are not very realistic, more or less the same as in 3.2. But with the added problem that in 4.0 they don't even do some very obvious and common tone sandhi's, which the 3.2 still do, like 不是, 不要, 一个, etc. In any case, they're both slightly robotic, the syllables don't flow as in real speech, sounding artificial, and it's clear it's a machine.

But what puzzles me is that, nowadays, when you hear any AI speak, even the free AI models, the voice is so realistic that it's almost indistinguishable from a real person speaking. And my Chinese friends tell me it's the same in Chinese, the TTS in AI models, even in the free Chinese AI models, is so good that they almost cannot tell if the audio is AI generated or it's a real person pronouncing really well.

I don't know if perhaps it's very difficult, or very costly, to have this kind of speech audio quality in Pleco too for the example sentences. But in case the Pleco team is considering it, I have to say that for me, this feature would be so useful that I'd be willing to pay for it, even if it was a relatively expensive add-on. Perhaps if other people agree with me it could be an option for Pleco in the future.


History

For me, History is very important, as it's the most used part of the app in my case. Once I've encountered words in a certain context and looked them up, I don't search for those words again. I just go to History to review the words from that context, repeatedly over several weeks or months. As I said above, reviewing is something you do many more times than searching for words.

Also, in History you usually have all the words from a certain context together, and their related words, as you searched them at the same time, and I actually leave them on my History list purposefully, because they should be reviewed together. For example, words with similar meanings, shared radicals or components, similar characters, similar pronunciation, related context, etc, it's very useful to leave them together on your History list, to review together.

I don't use flashcards of spaced repetition, because for me just scrolling through History is a simple, quick and accessible way of reviewing the words from the last week, or last month, or from a certain text I read recently, etc.

You can also test yourself using History, scrolling down one word at a time, so you see each word you're testing yourself on, right at the edge of the screen. If you scroll down, you see the Chinese first and you can test yourself from Chinese to English. But if you scroll up, you see the English definition first, so you can test yourself from English to Chinese which is the most challenging and useful type of test.

The problem with History in 4.0 is that the words don't have the tab for the example sentences (at least last time I checked), and example sentences are very important. So it would be an improvement to have the example sentences tab as well when you review words in History.

One good thing of History in 4.0 is that it has a search bar, so you can easily locate a word in your History list and when you saw it. This is very useful, and a big addition, but I think it cannot make up for the other problems.

So I really wish Legacy and 3.2 also had a search bar for History. If it's not too complicated to add, it would be a big improvement for those users that will choose to stay using Legacy.


User Interface

In my opinion, the screen space is used better in 3.2 than in 4.0. I'm referring to the dictionary, history, and definitions areas, where you see the definitions and example sentences under each dictionary, as for me these are the most used parts of the app.

In 4.0, under the definitions tab, you have each dictionary's definitions and examples contained within a square bubble headed by the name of each dictionary. This means you have less space available on the sides and the lines for text have to be slightly shorter. As cell phones are used vertically, the space on the sides is particularly valuable to prevent the lines of text from getting shorter, which means you need more lines to fit the same length of text.

In 3.2 on the other hand, the separation between dictionaries is just a horizontal line, which maximizes the space available on the sides and you can fit more text in less lines.

Furthermore, the name of each dictionary in 3.2 is displayed as the initials in small letters on the top right corner below the horizontal lines separating each dictionary, using minimal space. This size and placement is better, I think, because it's enough to serve its purpose well, as it's not the main piece of information on the page.

In 4.0, the name of each dictionary heads each of the square bubbles containing the definitions and sentences, which means the name of the dictionary uses a whole line for itself. Besides, it's in big letters; Even when I have my font settings for the definition body set to "large", the name of each dictionary will still show in slightly bigger letters than the definition body, using more space unnecessarily.

This means the overall view of this screen in 3.2 feels more spacious and clearer than in 4.0, with more space for the important things.


Anyway, this is just my opinion, based on the way I use the app, which probably is different for every person.

Even though I'm not planning to switch to 4.0 for the time being, I'd still like to thank and congratulate the Pleco team for making such a great app (including the Pleco dictionary!), probably the most essential app for students of Chinese and the most used app on my iPhone in the last two years.
 
Thanks!

Audio: the lack of sandhi transformations is a bug on our end, thanks for the reminder about that. As far as voice quality: basically it boils down to the fact that we need our TTS to be a) instant (even on older devices) b) offline and c) support character position highlighting; the TTS capabilities available to us lag by a bit thanks to those requirements, but we do expect that there will be new models that meet our needs soon - whether from our current provider or another one - and will be happy to switch when they do.

History: that's also a bug, will be fixed in our next beta update. We aren't currently planning to backport any major features from 4.0 to Legacy, we're more receptive to adding new options to 4.0 to fix gaps with Legacy once we have more feedback.

UI: this basically boils down to two things: making our app easier for new users to understand - the PLC / CC / etc were completely meaningless to people who didn't know our app well - and creating more visual spacing between dictionary entries. Also as a practical matter we've found that most lines of text in most dictionaries fit within the margins without wrapping, so it's not wasting much space. It's actually a better fit aesthetically for the new iOS 26/27 UI than the older 15/16/17/18 one, so we're quite pleased with how it looks in our new iOS 26 redesign.

That being said, though, the definition screen layout has generated a LOT of complaints and we actually added a 'compact mode' option to restore something like the old layout; we had to temporarily take it out due to some iOS 26 nonsense, but are in the process of restoring it, so that should be available again soon.
 
Just a suggestion:

Would it be possible to relax those 3 requirements?

(a)
I don't know how long the delay would be, but let's say users had to wait 2 or 3 seconds from the moment they tap the audio icon until they hear the sentence. For me, it would be totally worth it if I was to hear the sentence with "near native speaker" quality as the current AI models sound.

In fact, I think the ideal way to practice sentences, is to read the sentence yourself out loud, then hear it from a speaker, then "parrot" or emulate the speaker, especially if it's a native or near native speaker.

When doing this, you're actually practicing everything at the same time: visualizing the meaning of the word, while you say it, and hear it, which helps memorizing it. Hearing it in the context of the sentence, so you naturally learn its usage, collocations, grammar, etc. And pronouncing it yourself, right after a native speaker, the most natural way (toddler-like imitation) to learn a language.

And it's even better if you can repeat this two or three times per sentence, at least, while you study.

So when practicing this way, having those 2 or 3 seconds of silence since you tap the icon, is not a big problem at all, because you're going to be reading the sentence out loud yourself too, and that silence can be very conveniently used.

The rationale becomes even more valid if we compare it with the only alternative to practice sentences while listening to native speakers:

Let's say you pick a video on Youtube with its subtitles. The speaker is going to be talking continually, which forces you to use the pause button and tap to rewind. The minimum interval is 10 seconds, then 20 seconds, 30 seconds, etc. So tapping rewind will usually force you to wait more than 2 or 3 seconds until you reach the beginning of the sentence again, perhaps 5 or 10 seconds.

And they are not even seconds of silence that you can conveniently use to practice reading the sentence yourself, because the speaker is talking continually and he'll be interrupting you with other sentences.

Even if you choose one of those videos especially made for students where there are only example sentences separated by silence, the accuracy of the cursor in the progress bar of a video is such that you'll usually have to wait at least 5 or 10 seconds when rewinding.

And that's if you accept to listen to the sentences only in the order provided, one by one. It'll be even slower if you try to locate a specific sentence in the video.

Besides, the sentences in those instructional videos are usually randomly selected. Whereas Pleco's example sentences contain the word or expression that you're specifically trying to learn at that moment, and they have been especially selected by the dictionaries to learn that specific word, with all its senses and typical patterns of usage, which makes Pleco's sentences more useful and practical.

So basically, practicing listening/reading/repeating sentences is such an efficient way to learn a language that I would not mind at all about having that 2 or 3 second delay, or even more, 4 or 5 seconds, for me it would be totally worth it.

Just as I said previously, that I wouldn't mind at all if it had to be a relatively expensive add-on in Pleco, I'd gladly pay for it. ;)


(b)
The offline requirement seems even less important. Even if you don't live in a big city, you're probably online 99% of the time, and it's not like you're usually going to go study Chinese in the middle of a forest. So relaxing this requirement to get native speaker audio quality 99% of the time (or more) would make a lot of sense to me.


(c)
The character position highlighting requirement could be relaxed too, as it's not as important as having native speaker audio quality, for all the reasons I stated above.

And in fact, the highlighting sometimes can even be a little distracting. Because, when I'm reading a sentence as I listen to it's pronunciation, I'm usually trying to focus on the characters themselves, it's main components and radicals. So in that split of a second that I focus on each character as I follow along the sentence, the highlighting makes that sudden change where the background goes dark and the character strokes go from dark to white, precisely at that moment when you're focusing on it, so it can even make the reading a little less clear.

But even if most people didn't find the highlighting distracting, I still don't think it is such an important requirement, because in the average medium length sentence, it's fairly easy to visually locate and follow the characters as the speaker pronounces them.

Moreover, not having characters highlighted is the same way we listen and read text pretty much everywhere else. When you watch most types of Chinese content, podcasts, series, etc, even videos for language students, you'll usually have to read the subtitles without any highlighting, and it's not that difficult to follow along.

So it's a feature that we're very used to not having without being a problem.

Anyway, so in summary, that's why I think relaxing those three requirements would be a really beneficial trade-off, as having native speaker audio quality (or near native speaker as most AI models) would be much more valuable and useful to study Chinese.
 
Sorry, there's also a fourth issue: most of these engines don't allow us to override their pinyin to ensure that it matches ours. So we'd have the situation where the audio doesn't match our displayed pronunciation for an example sentence.

That being said, on Android devices you can install whatever third-party TTS engine you like and Pleco will happily let you choose it as the default Chinese TTS engine in Pleco, so a lot of people on Android are already getting state-of-the-art TTS (with those tradeoffs) in the app that way. I expect that Apple will either add their own support for third-party TTS engines or update theirs to match the quality of those soon; in the meantime, if there's a built-in Apple voice that you prefer over ours in quality you can use it in Pleco 4.0 via Settings / Audio Settings / Customize System TTS voices.

I would be reluctant to accept this set of tradeoffs in our default TTS - frankly there are a lot of other users for whom they would be a more serious problem than they are for you - but I'd certainly be open to selling an alternate TTS with a different set of tradeoffs at some point if an ideal replacement doesn't emerge soon. Given the speed of evolution in AI, however, it's quite possible that 6 months from now there'll be an open-source Chinese TTS that meets all of our requirements, so I'm a little reluctant to sell people something now that's going to be outdated and deficient soon.
 
Back
Top