VanModers114/East_Frisian_TTS

East Frisian Low Saxon Piper voice

This repository contains a single-speaker Piper ONNX text-to-speech model for East Frisian Low Saxon (frs). It was fine-tuned from the German Piper de_DE/thorsten/medium checkpoint and exported at 22,050 Hz.

Intended use

The model is intended for East Frisian language access, demonstrations, and research. Generated speech is synthetic. Do not present it as a recording of a real person, use it for impersonation, or rely on it where pronunciation errors could cause harm.

Limitations

• Pronunciation, prosody, numbers, abbreviations, and names can be incorrect.
• Coverage is limited by the training recordings and text normalization.
• The voice may inherit characteristics of the source checkpoint and training speakers.
• This model alone does not watermark audio. The accompanying Space applies an AudioSeal watermark and WAV provenance metadata after inference.

Files

• oostfraeisk.onnx: Piper inference model
• oostfraeisk.onnx.json: Piper voice and phoneme configuration

No license is asserted by this card. Users must verify that their intended use is permitted by the model, source checkpoint, training-data, and dependency licenses.