Character Mine Forum

Certain TTS voices don't always properly recognize words where a character is pronounced with the 5th tone, for example.

Just a "report" button might be useful to find out which TTS voices people are using and which questions people believe it's not reading properly. Some combinations of TTS voice and question could be confirmed correct.

I've run into some cases where the entire syllable is different for some reason. Perhaps something other than the TTS voice went wrong those times.

Whatever the cause, blacklisting the problem questions for each voice would make the whole experience flawless whatever voices someone decided to add.


Another thing would be an option for each TTS voice indicating whether not it should use pinyin input, because some TTS voices do support numbered pinyin input. Others have dictionaries where any format of pinyin could be added.

I actually ran into these problems with my userscript modification of Ma or Ma. I stopped using it because the TTS reading of the characters was too often different than the recordings.

I had been trying to find an unambiguous character for every syllable since the TTS voices don't seem to care what the meaning is when they're only reading a single word. This way I could have pinyin input of sorts for the voices that don't otherwise support it.

There's more I could think of doing with TTS voices but a I feel like a simple question blacklist would achieve the most for the least effort. Any work beyond that only helps gain back some of the edge cases.

Looking under \Documents\Character Mine Data\ToneQuestionUserData.xml I found something promising: I guess if I can figure out which question number my TTS voices can't read properly, I can just set <Banned> to true? Although, I'm guessing that would ban it for every voice even if only one of my voices is having trouble with it, but that's good enough for me.

Thank you for storing the settings in a human readable format.

Hi Youda,

Yes, setting "Banned" to true would have that effect, but as per your previous suggestion I have just published a version which features a Ban button, so it shouldn't be necessary. The reason the field is there is because the program was already automatically banning questions in cases where they provoked a runtime error, so it wasn't hard to add a button to do the same thing.

If and when there are any breakthroughs with TTS that resolve any of these problem sounds, for example the neutral tone which you mention, I may do a one-off reset of that field to re-include previously banned questions, but I will only do this when it is clear that the issues are really resolved.

Thank you for granting my wish so quickly. This makes things so much more convenient compared to having to break my flow and search for the question number to note down. It does still re-test the banned question until I get it right, but I don't mind that.