Chinese TTS

[jamesquek]

Mike,

Did you ever solicit for opinions on a Chinese tts engine. I think you did. If you did, it must be to replace the audio part of Plecodict. Any update on this? Personally, I think the paid audio feature is a bit of a letdown: limited prerendered multiple character words. Those not in the pre rendered list are not natural in their rendering. The interval between characters are too long.

 

[mikelove]

Should be in the Big Update. But we licensed AFAIK the best one available and it still can't match prerecorded audio for quality when the prerecorded audio covers a word, so I think there's value in both features.

 

[jamesquek]

Sounds like you're keeping the prerecorded audio feature and offering the TTS as another feature. Will this new TTS feature read the example sentences?

If you are keeping the prerecorded audio feature, can you improve on it by shaving the pause between each syllable in multi-syllabic words that are not prerecorded? With minimised pauses, I don't see why it could not be as natural sounding as the prerecorded ones.

Also, the neutral tones by themselves need reworking. They all sound like the first tone. If you compare the prerecorded audio of 扁担 entry and 扁担没扎，两头打塌 entry, you will see what I mean.

 

[mikelove]

Sounds like you're keeping the prerecorded audio feature and offering the TTS as another feature. Will this new TTS feature read the example sentences?

Yes.

If you are keeping the prerecorded audio feature, can you improve on it by shaving the pause between each syllable in multi-syllabic words that are not prerecorded? With minimised pauses, I don't see why it could not be as natural sounding as the prerecorded ones.

It's more than that - matching pitches, for example - and somewhat beside the point if we're also offering TTS; we do want to improve the quality of some of the recordings, though.

Also, the neutral tones by themselves need reworking. They all sound like the first tone. If you compare the prerecorded audio of 扁担 entry and 扁担没扎，两头打塌 entry, you will see what I mean.

Yes, we weren't able to collect good neutral tone audio for every word - again, an area to improve the recordings.

 

[Joey]

Will the TTS allow us to type/copy the characters into an entry box, and then the program could pronounce it all at once? For example, I can build sentences and hear it spoken for learning Chinese. Would it also be possible to generate the mp3 for it to save it for playback later on another device?

 

[mikelove]

Will the TTS allow us to type/copy the characters into an entry box, and then the program could pronounce it all at once? For example, I can build sentences and hear it spoken for learning Chinese.

Yes.

Would it also be possible to generate the mp3 for it to save it for playback later on another device?

Not currently possible for licensing reasons, sadly.

 

[Joey]

Can you make it do pinyin entry too? The Mp3 bit is for convenience but can be managed.

 

[mikelove]

Can you make it do pinyin entry too? The Mp3 bit is for convenience but can be managed.

Possible, but I'm not quite sure what the use case would be - is there a particular sort of text that you only have in Pinyin format?

 

[Joey]

I use it as a learning tool - and in my lessons I get sentences in pin yin as well as the original Chinese. Sometimes it is faster to just enter the sentence as Zhe4 ge4 gong1 ju4 hen3 you3 yong4 rather than looking up every single character to enter it in.