"Oh, Tannenbaum" artificially sung with a self-developed but AI-free TTS implementation.

This is "Oh, Tannenbaum" synthesized with my own self-developed still-unnamed text-to-speech and text-to-sing engine within my own self-developed DAW called klangknecht. My TTS Engine uses the MBROLA algorithm as a greatly improved from-scratch reimplementation in Object Pascal with more modern DSP algorithms (allowing for higher sample rates and bit depths, e.g. 48kHz 32-bit FP), so it is not the original MBROLA. Fortunately, the patent for the MBROLA algorithm expired a few years ago. My Object Pascal native TTS engine can therefore also use MBROLA voice databases, even for singing vocals, but it can also be purely synthetic with a KLATT-like synthesis model without any MBROLA voice database. It also contains two fairly complete txt2pho implementations of its own, one for German and one for English, which in my tests deliver significantly better results than, for example, using espeak(-ng) as txt2pho for the original MBROLA, also with regard to the speech melody. And this is also an extremely good example against software patents. Because this is actually 90s technology without all the modern new-fangled overhyped AI stuff, but where software patents have probably prevented it from being developed further up until the MBROLA algorithm patent has expired, in the way that I have now done it.
#ohtannenbaum #weihnachtsmusik #christmasmusic #delphi #lazarus #freepascal #objectpascal #pascal #dsp #daw #speechsynthesis #singing #mbrola

[@ika2318]

Very nice to see more videos of this project. Impressive as awalys. Great work!

[@au_trax]

Impressive - great to see a new video of your DAW; do you still work on Klangknecht?