A Pianist introduced us, and in our first meeting we discussed using a Musical Keyboard as a tool for Communication, “I type, u hear; u type, I hear.” The path was not clear, but we knew we needed Speech Synthesis with Musical (Prosodic) qualities.

We started work in a few days time. We worked on full Koran reading by Hossary. He recited the Koran observing [Arabic] Musical Scales, and a Tempo. He observed all the Musical rules, but did not sing. So his Phonemes should be good enough for our project. We transliterated the whole Koran text (using Hans Wehr), worked on Frequencies (Butterworth), Neural Networks (Keras and Random Forest), Amplitudes… etc. We exhausted two years ad nauseam on Mapping Phonemes. No useful results.

Then we created a “recipe” analogous to [Music] Sequencing, and we got 99% of the phonemes mapped. Then Synthesis started, and it took us another two years of work. But then we discovered that the foundations we used for Mapping limited our options on Synthesis.

However, the results we got (now on the website) are satisfactory, though not perfect by any means. Also, the 99% accuracy of phonemes meant there will be a mistake every 100 letters, and that in our opinion, is significantly high. So, we need to rewrite the whole program from scratch, and use the right foundations (that we think are suitable), and we expect to get 100% accurate results, with several options on Synthesis. Those options can easily reproduce Koran recital (obsering the rules), as well as the option of producing speech with Musical qualities. The rewriting of the programs will take several months, not years. But the results will be marked improvement in speech qualities.

We learned a lot in those 4+ years, and we look forward to much more. Please be patient, this project is a lifetime journey, and we enjoy what we do. It’s also worth mentioning that this work started on a Friendly basis.

We are two Friends working together, and we enjoy our work as much as we enjoy our Friendship…

Hope u like it…

Hans Boden