Monthly archives: March, 2016

Moresampler 0.6.0 Release

(Apr. 18, 2016 - Moresampler 0.6.3 is released!)

The highlight of this version is a novel set of flags: Moresampler Extension Flags. They are extremely helpful in adjusting timbre and creating many kinds of voice effect. All MEFs differ from other UTAU flags in that they have two characters in their names starting with 'M':

  • Mt: tenseness; range: [-100, 100]; default: 0
  • Mb: breathiness, setting to 100 leads to whisper; range: [-100, 100]; default: 0
  • Md: dryness; range: [-100, 100]; default: 0
  • Mo: openness; range: [-100, 100]; default: 0
  • ME: emphasis/de-emphasis formants; range: [-100, 100]; default: 0
  • Mm: model interpolation, which means to mix between the classical speech model used before version 0.3.0 and the cutting-edge speech model after version 0.3.0; range: [0, 100]; default: 100 (the new model)
  • Ms: stabilization - fix popping/clicking/growling sounds which may appear in some breathy voicebanks; range: [0, 10] (integer); default: 0; note: this flag lowers synthesis speed

Moresampler 0.5.0 Release

Faster than everything - this is Moresampler 0.5.0 in one sentence.

You may notice that 0.4.x versions got skipped. That's because the optimization and new features are pushing Moresampler toward 1.0.0, and 0.5.0 is a milestone in the halfway! Moreover, a 64-bit version is also included from now on.

If you are really interested in details, here's a list of what have been optimized:

  • Switch from GCC to MSVC (Microsoft's C/C++ compiler)
  • Improve file structure to speed up reading/writing
  • Avoid frequent memory allocations and frees
  • Skip irrelevant parts in .llsm data files when loading
  • Support multi-threaded synthesis based on OpenMP
  • Reduce FFT size by exploiting properties of energy signal
  • Reduce .llsm file size by 300% (with imperceptible quality loss)