New Android Beta 3.2.95

mikelove

皇帝
Staff member
Just submitted this to Google, can sign up for betas here to get it automatically once it's out.

This mostly addresses some long-simmering bugs - particularly Screen OCR permissions being wonky on Android 14 - but we also brought over two improvements from 4.0 beta on iOS: the option for an alternate OCR engine (in this case Google's MLKit instead of Apple's Vision), and, more importantly, our lovely new neural network TTS voices. (it also drops support for 32-bit CPUs, as, regrettably, those are not compatible with the new TTS)
  • Updated to a new, neural-network-based version of our paid text-to-speech engines, which should greatly improve audio quality and greatly reduce file size. (we are no longer offering 'enhanced' versions because the new regular voices sound much better than those do)
  • On Android 14 and later, switched Screen OCR to capture screenshots through the Accessibility framework (as Screen Reader uses) rather than Media Projection, which should eliminate a lot of annoying alerts and odd behavior related to using Screen OCR on 14/15. (it also puts us into better compliance with Google's new foreground service restrictions)
  • Rejiggered our shortcuts system to let you create a shortcut icon for any command you can add to the shortcut menu this can, however, only be done through settings. (we also added 'do screen reader' and 'do screen ocr' shortcuts since we offered icons for them previously but did not previously support them in the shortcut menu)
  • Fixed a bug that flipped paused images from New OCR upside-down after returning from another screen.
  • Fixed (we hope) some bar positioning bugs with Screen OCR, especially in cases where the navigation bar is hidden (e.g. scrolling down a PDF).
  • Fixed some bugs on foldable screens (finally got an emulator for one) the app should now force itself into a phone UI with a portrait oriented foldable screen.
  • Fixed a bug that could cause page/chapter ebooks to lose their place in a chapter when rotating the screen rapidly.
  • Fixed a bug that could prevent single-long-scrolling reader documents from scrolling the entire highlighted range into view if it's multiple lines long.
  • Fixed the 'save updated entry' icon in flashcard tests being a + instead of a down arrow.
  • Made the "Pleco" system selection menu command follow startup clipboard search behavior in deciding whether to bring up that text in the search box or the document reader.
  • Improved the reliability of screens updating after a simplified/traditional toggle. (both in sidebar and in settings)
  • Added back experimental handwriting option, plus a new experimental option for still OCR (including Screen OCR) the latter has some accuracy / positioning advantages over or current algorithm but does a very poor job of mapping character positions on the screen, so with it enabled we hide the character overlay.
 

kaine42

Member
Hello, This is a great update, I just can't find the option to change to the alternative google ocr engine.
I only get the choice of classic or new experimental engine in the ocr settings, I don't know if it's this one ? But I already had this in 3.2.94 so I don't think it's this one or if it's in another setting ?
Thanks !
 

mikelove

皇帝
Staff member
Thanks!

The option should be at the top of the 'still image' section - 'still/screen OCR algorithm.' (we aren't yet offering the option to use it in live)
 
What's different with the experimental handwriting feature? I switched it on, and in the process discovered that there is a cursive character database and an option to have common subsequent characters suggested after a character is recognized/selected to turn on. I definitely can immediately see the difference with the latter!

Also really cool to see 4.0 enter beta, even if it's only on iOS at the moment.
 

mikelove

皇帝
Staff member
That latter one is actually supported even without experimental handwriting, it's been around for a while. Do you notice any difference in recognition with the experimental one? (it's genuinely an experiment, if nobody sees any benefits then we'll remove it again rather than waste app binary space on it)
 

mikelove

皇帝
Staff member
FWIW, we just released another - hopefully final - beta of this that updates all of our third-party libraries to support 16kb page sizes, future-proofing Pleco for new devices with larger amounts of internal RAM. Sorry for the wait; assuming this checks out we should be able to release a final version pretty soon.
 

Dkhr77

Member
First want to say, Pleco is a fantastic piece of an app, you're doing great!

Version 3.2.95 introduced a wrong audio for the 滤镜 entry,it speaks it like biăojìng in a strange way, however 滤 sounds like it should be.
 

mikelove

皇帝
Staff member
Thanks - it seems like there's a bug in how the new 16kb-friendly version of our TTS engine processes ü's, I've just uploaded a fix to Google Play so the problem should go away once they approve it.
 

Dkhr77

Member
Thanks Mike, that fixed it.

However i have now experienced another bug. Pleco will crash for the 三天打鱼两天晒网 entry if
  • I open the entry and click the loudspeaker icon to play audio next to the pinyin
  • I search the entry in the dictionary and choose "play audio" from the context menu after long pressing the entry
  • it also crashed my test in review with play audio on reveal
However, it will not crash if you open the entry and select the complete headword and then use the loudspeaker icon in the top menu (right beside copy to clipboard).
 

mikelove

皇帝
Staff member
Thanks, we'd pushed a fix for this already but apparently it didn't use the correct library file so I've uploaded another one now.
 

Pismo

Member
Does the speech on this one sound strange to anyone else? The female voice sounds like a different first syllable on my devices.

晚安 (wǎn'ān, good night)
 

Pismo

Member
Oh man, I'm so new to this I'm not sure how to answer that. Let's see...Settings/Audio/Headword Audio/Preferred gender/Alternating. Under "Preferred Gender (Cantonese) is also Alternating. There's a check box selected for "Use Cantonese TTS if no recording". Oh, under Sentence Audio it says TTS cycle voices and there is abox checked for System Madarin female.

So, my guess is either the female Cantonese or female Madarin. Maybe you'll know for sure with the above?

Also, to me she sounds like she's saying "bah an" or "bal an".
 

mikelove

皇帝
Staff member
OK, so it sounds like it's Google's system voice. If you go into Settings, Audio, System TTS Settings you might be able to switch to a different default Chinese voice in there - does that help matters any?
 

Pismo

Member
Unfortunately not. I can't even tell if it's a recording or a synthesized voice. I think I'd have to sit down and take notes on each change as I tried to find the combination that exposed where the particular voice is and it's not worth the time. I'll just live with it since at least the male voice seems fine.
 

mikelove

皇帝
Staff member
You can set it to use only the male voice, or you can set "system TTS" to "none" and then it will just rely on those male recordings and their female counterpart.
 

Pismo

Member
I did that and the first time I click on the speaker I hear the male voice, second time, female voice and she still says "băl,ān" lol. Like I say, I can live with. I was just being a little too AR!
 
Top