New TTS voices Lulu and Hui

[siamesefightingfish]

This add-on is simply amazing. The TTS readers read sentences at the speed of a native speaker. Faster in fact, I slow them down to 80% while I ingest the new vocabulary. Awesome! I still prefer the recorded voices, but these are almost as good. Some Chinese learners prefer to focus on reading and writing, or, like myself, speaking and listening. So I'm a bit particular about the audio. Previous to installing the TTS voices I had a little gripe about the recorded voices sounding unnatural on certain words, like "German person" The recorded sound "German" and "person" were put together to sound slow and unnatural). After installing the TTS voices however, those certain words are now defaulted to the TTS reader (Hui is my favorite), which reads them at normal speed, so every word played now is at native speaker speed, and natural sounding. This is exactly how I would have set this up had I done the programming.
How it works: When you search a word in the dictionary, you can play it with the recorded voices (if you bought and installed them), then, if you have a dictionary with example sentences, a little speaker icon will appear in front of the sentence allowing you to play the whole sentence with the TTS voices. Finally, if you want a syllable by syllable play of the word, just click on it and the free "Audio Pronunciation" voices will play. In my configuration, I now have 6 possible voices speaking to me depending on how I want to hear a word or sentence. This completes the package.

 

[mikelove]

After installing the TTS voices however, those certain words are now defaulted to the TTS reader (Hui is my favorite), which reads them at normal speed, so every word played now is at native speaker speed, and natural sounding. This is exactly how I would have set this up had I done the programming.

Thanks! We came up with a number of ideas for alternate options for how this might work (e.g. length-based cutoffs, only using recordings for single syllables, only using TTS in something that contains enough punctuation to suggest that it's a full sentence, etc), and we can always implement those if we find that this approach causes problems for some people, but we're hoping it'll satisfy almost everyone.

Finally, if you want a syllable by syllable play of the word, just click on it and the free "Audio Pronunciation" voices will play.

That's actually not how it's supposed to work... when exactly is it doing this? Just for single syllables (which is expected) or for multi-syllable words as well?

 

[siamesefightingfish]

Just the syllables Mike. Was just pointing out that it still has that function, in combination with the others. Im really liking this setup. Xiexie!

 

[kyococo]

I tried Lulu on Tuttle with this entry:

按
àn
PREPOSITION according to, in accordance with

> ∼計劃, 這座工廠將在明年四月建成。

> 我一定∼你說的做。


Here is my quick feedback:
- speaker icon not very responsive, 2 sec delay making you wonder if you pressed it correctly, especially as the icon is small on big screens like the note's
- doesn't pronounce the entry word (represented by the tilde) making the phrase awkward
- well if only... Much worse, skip many words or put blanks...
On this entry it read:座工在明年四月建成
我一定你__pause___的做。

it's still experimental, but then maybe you should make it a beta?

 

[gato]

It seems that Hui is much better than Lulu. Would it reduce user issues if you removed Lulu as an option?

 

[mikelove]

- speaker icon not very responsive, 2 sec delay making you wonder if you pressed it correctly, especially as the icon is small on big screens like the note's

What sort of device are you using this on now?

- doesn't pronounce the entry word (represented by the tilde) making the phrase awkward

Well that's an odd one, we carefully went in and fixed all of the ~s in Tuttle - have you configured Pleco to replace them with ~s specifically?

- well if only... Much worse, skip many words or put blanks...
On this entry it read:座工在明年四月建成
我一定你__pause___的做。

Probably related to the icon pause, but if we can reproduce it here it should be addressable.

it's still experimental, but then maybe you should make it a beta?

To be honest we view "experimental" as a way of saying "beta" that's easier for non-technical people to understand - did it not come off that way to you?

It seems that Hui is much better than Lulu. Would it reduce user issues if you removed Lulu as an option?

A lot of people seem to like both, actually - prefer the sound of Lulu - and there are a few devices on which Hui for whatever reason fails to run at all. However we certainly might end up making Lulu "experimental" for a longer time than Hui, and it's 50-50 whether we release Lulu at all on iOS (where Apple is generally less tolerant of "experimental" things).

 

[Michaels1980]

Both Hui and Lulu are working well for me on my Galaxy S3 and Nexus 7. I found Hui slightly too fast so I slowed her down to 80%. I'm not sure which I prefer yet - some sentences sound better with Hui and some with Lulu - but overall I think TTS is excellent.

 

[kyococo]

- speaker icon not very responsive, 2 sec delay making you wonder if you pressed it correctly, especially as the icon is small on big screens like the note's

What sort of device are you using this on now?

As I said a galaxy note
I switched the setting to prefer Lulu and no more delay... Seems like it takes him 2 sec to check if HUI is installed

Well that's an odd one, we carefully went in and fixed all of the ~s in Tuttle - have you configured Pleco to replace them with ~s specifically?

That's the default setting... I checked the "do not replace head word" and I no longer have tildes which solve the issue at least for Tuttle

Probably related to the icon pause, but if we can reproduce it here it should be addressable.

Many many words seem missing as they are systematically skipped in other examples...

To be honest we view "experimental" as a way of saying "beta" that's easier for non-technical people to understand - did it not come off that way to you?

I was saying that it could be free at this stage of development... Although it is cheap, you would also have more feedback.

A lot of people seem to like both, actually - prefer the sound of Lulu - and there are a few devices on which Hui for whatever reason fails to run at all. However we certainly might end up making Lulu "experimental" for a longer time than Hui, and it's 50-50 whether we release Lulu at all on iOS (where Apple is generally less tolerant of "experimental" things).

I also prefer Lulu... Please make HUI the forever experimental one

 

[mikelove]

Many many words seem missing as they are systematically skipped in other examples...

Hmm... could be that it's not doing well with fanti; I believe it's supposed to support both but it may be working with a smaller wordlist. Does it work better if you temporarily switch into jianti mode?

I was saying that it could be free at this stage of development... Although it is cheap, you would also have more feedback.

Not possible for licensing reasons, sadly.

 

[leoninto]

Can we purchases these voices for use on both Android and iPhone (later)? Like we're able to purchase on iTunes, then convert the license to Android to use on both? If we can, I'm going to buy both right away.

 

[kyococo]

Hmm... could be that it's not doing well with fanti; I believe it's supposed to support both but it may be working with a smaller wordlist. Does it work better if you temporarily switch into jianti mode?

Indeed it's much better in simplified.

After some testing only found an issue with this phrase:

> 小河两岸柳树成行。
Xiǎohé liǎng'àn liǔshù chéngháng.
On either bank of the stream stand rows of willow trees.

She read 成行 as:
1chéngxíng*
V. embark on a journey

How is she supposed to manage heteronyms anyway?

 

[mikelove]

Can we purchases these voices for use on both Android and iPhone (later)? Like we're able to purchase on iTunes, then convert the license to Android to use on both? If we can, I'm going to buy both right away.

Unfortunately no, Apple doesn't let us sell iOS add-ons outside of their store. But our policy is that if you buy something on Android through our online store (rather than as an in-app purchase), then buy the same thing on iOS within one year, you can send us proof of purchase from that iOS version and we'll cancel / refund your Android purchase. (most of the time we're happy to offer this with in-app purchases too, but we can't guarantee that because Google could theoretically decide at any time to block us from doing so - with our own store there's nobody in a position to tell us what we can't do)

One added wrinkle in this case, though, is that we haven't yet decided whether to offer both Lulu and Hui on iOS - we may end up with only one of them.

Indeed it's much better in simplified.

Great! Then for a future update I guess we'll start feeding her simplified versions of things if you've got Pleco in traditional mode.

How is she supposed to manage heteronyms anyway?

Mostly from context, though Hui's rules for doing that seem to be a bit more sophisticated at the moment.

 

[leoninto]

Can we purchases these voices for use on both Android and iPhone (later)? Like we're able to purchase on iTunes, then convert the license to Android to use on both? If we can, I'm going to buy both right away.

Unfortunately no, Apple doesn't let us sell iOS add-ons outside of their store. But our policy is that if you buy something on Android through our online store (rather than as an in-app purchase), then buy the same thing on iOS within one year, you can send us proof of purchase from that iOS version and we'll cancel / refund your Android purchase. (most of the time we're happy to offer this with in-app purchases too, but we can't guarantee that because Google could theoretically decide at any time to block us from doing so - with our own store there's nobody in a position to tell us what we can't do)

If we refund our Android purchase, can we still use it on Android, or essentially we're only allowed to "switch" to iOS later?

When do you think you'll know if you're offering both on iOS?

 

[mikelove]

Actually I have to correct this - we haven't technically gotten sign-off from the makers of the TTS add-ons on platform transfers yet, though we'll be happy to do the refunding business at least.

If we refund our Android purchase, can we still use it on Android, or essentially we're only allowed to "switch" to iOS later?

Assuming that we do get the OK to platform-transfer them, you'd be able to use them on both, just as with any other add-ons - Google doesn't block us from selling stuff outside of their store like Apple does.

 

[leoninto]

Is it just me or are the samples not working? I get an error saying the audio sample file appears to be temp. inaccessible. It's been like that for more than a day.

I know my firewall isn't blocking Pleco because I can click the Pre-Download and it downloads quickly.

It wouldn't be hosted on a server blocked by AdBlock/AdAway for Android, would it?

I remember one of them sounding nicer than the other, and want to purchase one now to play with.

 

[mikelove]

Is it just me or are the samples not working? I get an error saying the audio sample file appears to be temp. inaccessible. It's been like that for more than a day.

Seems to be working fine here...

It wouldn't be hosted on a server blocked by AdBlock/AdAway for Android, would it?

I don't believe so, no - it's the same server where we host our other data files, so if you can download the module itself then you should be able to download the audio.

Direct download links for the samples are here:

Hui

Lulu

 

[phaneron]

Just wanted to say I'm loving the new TTS voices, they're really helpful. By and large they work perfectly, the only major glitch being, as noted above, that Lulu systematically skips every fanti character that does not also occur as a jianti character. eg. with 繁體 she'll read the first character but skip the second, with 語言 she'll skip the first and read the second.

I prefer traditional characters, but also quite like the Lulu voice, so a little torn at the moment! Hopefully this is something that's not too hard to fix.

Mark

 

[mikelove]

I prefer traditional characters, but also quite like the Lulu voice, so a little torn at the moment! Hopefully this is something that's not too hard to fix.

Not too hard, no - hopefully fixed in another update or two.

 

[leoninto]

How do we use the new voices to read text from the Reader? I only got around to trying out the reader with tts and thought if I copied in a piece of text, I could simply have the tts read everything out?

I read through the Reader section in the new 2.4 manual, but didn't see anything about it.

 

[mikelove]

How do we use the new voices to read text from the Reader? I only got around to trying out the reader with tts and thought if I copied in a piece of text, I could simply have the tts read everything out?

I read through the Reader section in the new 2.4 manual, but didn't see anything about it.

Has to be done one paragraph at a time at the moment; tap on a word to select it and tap-and-hold on the |-> button twice to expand it first to a sentence and then to a paragraph, then tap on the audio button to play it.