Free Software Virtual Singer
summary
Describes an open-source virtual singing system built from free components: a GNU GPL Spectral Modeling Synthesis (SMS) implementation combined with the real-time WORLD synthesis engine and an UTAU-compatible resampler. Covers a documented free voice-bank format plus resampling, pitch-shifting, segment-stretching and compressed spectral frames to make CD-quality voice data smaller than WAV, with working Japanese demos and planned English and German ports.
AI-written summary · about
chapters (12)
- 0:00Start
- 1:50The VOCALOID Software
- 3:30VOCALOID/MBROLA DSP
- 9:35eCantorix/QTau
- 14:58Using the Cepstrum in WORLD
- 17:38Using the Group Delay in WORLD
- 20:16Voice recording
- 22:47Voice labelling
- 27:24Replacing MBROLA/VOCALOID
- 28:55QTAU Editor
- 29:47Audio Examples
- 46:21Free Software Virtual Singer
related talks
51mThe Merger of Machine Learning, Music and Programming
54mIntegral scherzo: how machines compose and sing
30mSound Modelling - parametric methods and deep learning...
41mMaking Machines that Make Music
16mDemo: Functional Sound Design
27mScala Sampler for Functional Soundscapes
20mLive Coding With Crowdsourced Sounds And A Virtual Agent Companion (second session)
45mMusic Production under GNU/Linux