Are there any CV voice banks recorded in a manner similar to VCVs or CVVCs? To keep the file size down, why don't we record a string of kana slowly and break them apart. For example, we could record kakikukeko and duplicate them to get each syllable. Need romaji aliases? Duplicate the hiragana duplicates and change it to romaji. Also, a basic hiragana bank with no extras has about 104 samples. With the string kana way, we can have only 21 samples if I did the math right. 20 strings of five and one string of four. Or we can reduce the amount to 20 if one is a string of six. What do you think?