Making MAI pronounce plosives more clearly - right now P sounds like B

I’ve been looking at phonetics schemes and tried various solutions but I cannot get MAI to pronounce P in “party” correct. It sounds like Barty. Also, C in “care” sounds more like a G. I have not been able to crack the phonetics to make it sound right. Please if anyone can help, it will be most appreciated.

Similar sounds, especially in the case of a leading b or p, can sometimes be influenced by the pitch curve. Try different pitch curves leading into the note (ie rising or falling pitch during the p phoneme). p can sometimes also have more airflow than b, so perhaps try increasing the breathiness during the phoneme as well.

For c and g, try reducing the voicing and increasing breathiness at the moment of that phoneme. They’re the same mouth shape but without and with voicing respectively.

1 Like

I think you may try to move just these notes beginning with p and g slightly whether more to the right on the grid or just through the parameters

1 Like

Thanks a lot. I will give it a go. I am trying to recreate a demo to give it a new feel, and to try out Synthesizer V in an actual work situation. Solaria sounds a little too schooled and english :slight_smile: Mai sounds great and I am thinking of getting Eri, but I need to find out if the crossover language tool, is actually working pronouncement wise. Maybe a good dictionary can help. Right now I am looking at various phoneme charts. But I really appreciate your comment.

1 Like

There are voiced/unvoiced phonemes, such as b/p, d/t, g/k, z/s and zh/sh.

Since the voiced and unvoiced versions of the phoneme pairs have the same articulation, you can turn any of these voiced phonemes into the unvoiced equivalent by going to the Parameters panel, choosing Voicing, and removing the voicing under the voiced phoneme. In the example below, the b will be pronounced as a p, as in “pat”:


In your case, you can just de-voice the p so it’s pronounced correctly:



This is also interesting, thanks

I will definitely try this out today. Thanks a lot. Poeple in this forum are truly amazingly helpful. I will post my finding in here so others can benefit from your input and my mistakes :slight_smile:

1 Like

Excellent - thank you. I have been having the same issue with Sheena for some contexts, eventually using a different voice as I was unhappy with Sheena’s plosive sounds. I have since tried your suggestion with Sheena and for my example it worked very well. I also had an issue with ‘flowing fast’ sounding as ‘flowing vast’. The de-voicing didn’t seem to work as well for this: in this case another suggestion (I think also by you) was to precede the phoneme set with a closure (cl) - this improved the pronunciation too for this example.

I am interested to note that for my work with Sheena, the pronunciation issues only occur with some lyric-phoneme sequences, whereas pronunciation for others is OK. This makes me wonder whether it would be possible for Dreamtonics to improve Sheena’s pronunciations by refining the AI.

In note properties, setting Duration and Strength of the “P” phoneme to 100 will give it a stronger P sound as well.