Audio8-TTS-Realistic-0.6b GGUF

Run with audio.cpp using the existing audio8_tts family. No model-specific runtime extension is required.

Upstream model: Edge0/Audio8-TTS-Realistic-0.6b.

Text to speech

audiocpp_cli --family audio8_tts --task tts --backend cuda \
  --model audio8-tts-realistic-0.6b-q8_0.gguf \
  --text "Welcome to Audio Eight. It is wonderful to hear from you." \
  --out speech.wav

For BF16, replace the model filename with audio8-tts-realistic-0.6b-bf16.gguf. Use --backend vulkan or --backend cpu for those backends.

Voice cloning

Provide a clean reference recording and its exact transcript:

audiocpp_cli --family audio8_tts --task clon --backend cuda \
  --model audio8-tts-realistic-0.6b-q8_0.gguf \
  --voice-ref reference.wav \
  --request-option "reference_text=The exact words spoken in the reference recording." \
  --text "This sentence is spoken with the reference voice." \
  --out clone.wav

Both packages include the model, codec, tokenizer, and configuration. Output is mono 44.1 kHz audio.