etvoices

Repository to train and use piper-tts voices from Enemy Territory samples.

Requires

• sox
• podman
• an nvidia gpu (by default, CPU is also possible but slower, AFAIS no AMD) + https://wiki.archlinux.org/title/Podman#NVIDIA_GPUs

Usage

1. Unpack all Enemy Territory .pk3s into some temporary directory.
2. Move ./sound/vo/ from this temporary directory to your etvoices checkout ./vo-in/
3. Run ./0-cut-samples.sh
4. Run ./1-prepare-dataset.sh <team>
5. Run ./2-train-model.sh <team>
6. Run ./3-export-model.sh <team>
7. Run ./4-say.sh <team> <message>