Audio8 releases 0.6B ONNX INT4 TTS model: CPU-only, 11 languages, zero-shot voice cloning

airesearch12 · x · 2026-08-01

Audio8 TTS Preview 0.6B ONNX INT4 is now available, designed for low-resource CPU inference. The 0.6B-parameter model runs entirely on CPU with about 1GB memory usage during inference, supports multilingual TTS in 11 languages and zero-shot voice cloning. No PyTorch or Transformers required; includes CLI, streaming, HTTP/OpenAI-compatible APIs, and voice registration. Model files are on Hugging Face, code in the Audio8 TTS repo.

Original post →

More from Models

Models channel →