jamcoder

majical voice jamcoder

uv Virtual Environment and Dependency Installation

Install uv before continuing.

uv venv
uv pip install -r requirements.txt

Prerequisite: Phoneme Dataset

See data/README.md

Usage

[jamcoder]$ uv run python src/jamcoder.py --help
usage: jamcoder [-h] [-nc] [-w CROSSFADE_OVERLAP] [-nd] [-o OUTFILE] -u VOICE -s SENTENCE [SENTENCE ...] [-d]

options:
  -h, --help            show this help message and exit
  -nc, --no-crossfade
  -w, --crossfade-overlap CROSSFADE_OVERLAP
                        Sets the overlap of the crossfade, from 0.0 to 1.0. Higher means more overlap.
  -nd, --no-dual-similarity
  -o, --outfile OUTFILE
  -u, --voice VOICE
  -s, --sentence SENTENCE [SENTENCE ...]
  -d, --debug

Generating a sample

Lets say you have a dataset of TextGrids and wavs stored at data/james.

You can have this james voice synthesize a (plausible) sentence with:

$ uv run python src/jamcoder.py -u james -s hey whats up

This will generate synthesized speech to synth.wav in your present working directory.