Monthly Archives: April 2015

Bug Fixing – The Fun Never Ends

I’ve started running other songs through synSinger, which has uncovered another slew of bugs. Some of them were in the rules-based pronunciation code, and never came to light because the lookup dictionary caught most of the words. There are some … Continue reading

Posted in Uncategorized | Leave a comment

Humanizing

The logic for calculating the amplitude of voicing has been rewritten. It seems to work well at slow tempos, but not so well for faster tempos. So that will need some revisiting. The code for transitioning from pitch to pitch … Continue reading

Posted in Uncategorized | Leave a comment

Adding Some Jitter

To counter some of the robotic effect of the voice, various parameters – such as the pitch and formant – can be jittered by random numbers. Dennis Klatt suggested using three slowly moving sin() waves. Basically: sin(2*PI*12.7*tick) + sin(2*PI*7.1*tick) + … Continue reading

Posted in Uncategorized | Leave a comment

Working on Voice Quality

I’ve changed the code to use a simpler transition model, and got it mostly working again. The new model has default HELD and TRANSITION durations for each phoneme, and is a lot simpler to implement than the prior model. There … Continue reading

Posted in Uncategorized | Tagged , , , , | Leave a comment

Breaking code again

I’ve decided to simplify the model that synSinger uses for creating transitions. There are a couple of motivations: The current model is complex I don’t understand where all the values came from I don’t see them supported in analysis that … Continue reading

Posted in Uncategorized | Leave a comment

Using Samples

synSinger now uses samples for unvoiced – and a number of voiced – consonants. Additionally, it supports a number of sampled blends, such as ST and PS. While there’s quite a ways to go in terms of realism, the sampled … Continue reading

Posted in Uncategorized | Leave a comment