piper-sample-generator 3.2.0

Generate TTS audio samples for training wake word systems

Piper Sample Generator

Generate spoken audio samples using Piper for training a wake word system like openWakeWord or microWakeWord.

Supports normal Piper voices or a special generator that can mix speaker embeddings (English only).

Install

Create a virtual environment and install the requirements:

git clone https://github.com/rhasspy/piper-sample-generator.git
cd piper-sample-generator/

python3 -m venv .venv
source .venv/bin/activate
python3 -m pip install --upgrade pip
python3 -m pip install -e .

Piper Voices

Download one or more Piper voices (both the .onnx and .onnx.json files for each voice). Audio samples are available.

As an example, we'll download the U.S. English "lessac" voice in medium quality:

mkdir -p voices
wget -O voices/en_US-lessac-medium.onnx 'https://huggingface.co/rhasspy/piper-voices/resolve/main/en/en_US/lessac/medium/en_US-lessac-medium.onnx?download=true'
wget -O voices/en_US-lessac-medium.onnx.json 'https://huggingface.co/rhasspy/piper-voices/resolve/main/en/en_US/lessac/medium/en_US-lessac-medium.onnx.json?download=true'

Generate a small set of samples with the CLI:

python3 -m piper_sample_generator 'okay piper.' --model voices/en_US-lessac-medium.onnx --max-samples 10 --output-dir okay_piper/

Check the okay_piper/ directory for 10 WAV files (named 0.wav to 9.wav).

You can add multiple --model <voice> arguments to cycle between different voices when generating samples.

See --help for more options, including --length-scales (speaking speeds).

Generator

Download the LibriTTS-R generator (exported from checkpoint):

wget -O models/en-us-libritts-high.pt 'https://github.com/rhasspy/piper-sample-generator/releases/download/v2.0.0/en_US-libritts_r-medium.pt'

Generate a small set of samples with the CLI:

python3 -m piper_sample_generator 'okay piper.' --model models/en-us-libritts-high.pt --max-samples 10 --output-dir okay_piper/

Check the okay_piper/ directory for 10 WAV files (named 0.wav to 9.wav).

Generation can be much faster and more efficient if you have a GPU available and PyTorch is configured to use it. In this case, increase the batch size:

python3 -m piper_sample_generator 'okay piper.' --model models/en-us-libritts-high.pt --max-samples 100 --batch-size 10 --output-dir okay_piper/

On an NVidia 2080 Ti with 11GB, a batch size of 100 was possible (generating approximately 100 samples per second).

Setting --max-speakers to a value less than 904 (the number of speakers LibriTTS) is recommended. Because very few samples of later speakers were in the original dataset, using them can cause audio artifacts.

See --help for more options, including the --length-scales (speaking speeds) and --slerp-weights (speaker blending) which are cycled per batch.

Augmentation

Once you have samples generated, you can augment them using audiomentation:

python3 -m piper_sample_generator.augment --sample-rate 22050 okay_piper/ okay_piper_augmented/

This will do several things to each sample:

1. Randomly decrease the volume
   • The original samples are normalized, so different volume levels are needed
2. Randomly apply an impulse response using the files in piper_sample_generator/impulses/
   • Change the acoustics of the sample to sound like the speaker was in a room with echo or using a poor quality microphone
3. Resample to 16Khz for training (e.g., openWakeWord)