putao 0.4.1

Poor man's UTAU

putao: Poor man's UTAU 🍇

putao is a Python module that allows programmatic creation, editing, and saving of UTAU-style songs.

No more having to wrestle with encoding issues: all filenames and text files use UTF-8, re-encoded from Shift-JIS if necessary.

Originally, it was a joke inspired by Composite's Bad Apple script but is now an experimental attempt at creating a voice synthesizer in Python.

Why name it putao?

Because python + utau = putao. Conincidentally, it means 'grape' (葡萄) in Chinese.

How it works

putao uses UTAU-format voicebanks (with oto.ini files).

The key difference is that putao does not shell out to resampler.exe, or any other .exe compiled resampler. putao's resampler is written entierly in Python, thanks to pyworld.

(Which means it is cross-platform, as long as you have a C compiler for pyworld.)

The resampler's core logic is in model.py, which models UTAU elements as python classes. To create a new resampler, inherit from the Resampler class and override the abstract methods.

To save time, putao generates frequency files from pyworld for all the wavfiles in oto.ini, similar to .frq files generated by UTAU's resampler.exe. These analysis files are just numpy arrays saved in numpy's native format (hence the .npy extension).

Later on when rendering songs, the frequency files can be loaded back into the resampler.

NOTE: You have to use putao extract <voicebank zipfile> -t <folder> so the oto.ini and filenames can be read correctly by putao!

Making a song

.mml song files are written with an extended form of Music Macro Language, described below.

Check out some examples: to render them, do the following: (You need to supply your own UTAU voicebank, though. Make sure to extract with putao extract for utf8!).

cd examples
putao new <name>
putao render -s <name>.mml

where name is one of the .mml files, i.e bad_apple.

putao is CLI-only for now; creating a GUI to make songs is currently high priority, and it will be worked on once putao's API is stabilised.

MML Syntax

(Anything enclosed in '[]' is optional.)

lengths below may be one of '1', '2', '4', '8', '16', '32', '64' (1 -> whole note, 2 -> half note, 4 -> quarter note, 8 -> eighth note, etc.)

Core syntax:

(key)[accidental][length]
    Play a note.

    key must be one of the letters 'abcdefg',
    and accidental may be '+', '#' (sharp) or '-' (flat).
    i.e 'c2'

p[length] / r[length]
    Pause/rest playing notes.

o(octave)
    Change the octave of the notes.

    octave must be a integer.

> / <
    Shift current octave up or down.

l(length)
    Change the length of the notes.

t(tempo)
    Change the tempo of the notes.

    tempo must be an integer, in beats per minute.

Extended syntax:

#[comment]
    A comment.
    Comments can appear anywhere in a line (i.e after commands).
    All text between the '#' until the next newline is ignored.

@(trackname)
    Add any notes after this to trackname.
    Any notes without a specified track are implictly added to the 'global' track.

|(lyrics)
    Add lyrics to the current track, split by whitespace.
    Each split is a phoneme.

Todo

• Add midi support (WIP, 50%)
• GUI

Install

pip install putao

On Linux, you may have to install your distro's equivelent of the following packages:

• libsnffile1

i.e Debian:

sudo apt install libsndfile1

License

MIT.