To provide a voice synthesizing device, etc., which can easily secure a sufficient amount of data for voice synthesis and properly protect the right of those data.

A language processing part 1 takes a word analysis and a modification analysis of a sentence to generate a phonetic character string. A sound processing part 3 searches a phoneme dictionary storage part 4 for enciphered phoneme spectrum data showing phonemes in the phonetic character string and supplies them to a deciphering part 5-1, etc. The deciphering part 5-1, etc., deciphers the enciphered phoneme spectrum data with deciphering keys by bands and supplies the obtained phoneme spectrum data to the sound processing part 3. The sound processing part 3 determines the continuance and basic frequency pattern of phonemes and determines the waveform of the voice of the whole sentence. This voice is outputted from a voice output part 6. A phoneme spectrum data generation part 7 quantizes the spectrum of the data of the waveform of the phonemes almost to a formant frequency with high precision and stores it in a phoneme dictionary after electronic watermarking and ciphering.