NbAiLab/nb-tts-norwegian-voices

NbAiLab Norwegian Voices

This repo contains a two-voice Norwegian MOSS-TTS-Local fine-tune.

Training data:

• Dataset: NbAiLab/nb-tts-voices-2607
• Voices: female (speaker_id=M) and male (speaker_id=T)
• Total prepared data: 10,310 clips, about 19.08 hours

Published checkpoint source: checkpoint-epoch-19

Inference notes:

• Load with AutoProcessor.from_pretrained(..., trust_remote_code=True).
• Load with AutoModel.from_pretrained(..., trust_remote_code=True).
• The checkpoint uses the MOSS Local remote-code files bundled in this repo.
• Voice selection is best done with reference-audio prompting.