Audio8 releases 0.1B TTS with 11 languages and zero-shot cloning on CPU

alexcovo_eth · x · 2026-08-26

Audio8 released a 0.1B parameter TTS preview supporting 11 languages and zero-shot voice cloning. Available on Hugging Face as an ONNX INT8 build, it runs on CPU only (0.4 GB RAM) without PyTorch or CUDA dependencies. It outputs 44.1 kHz audio, features an OpenAI-compatible API, and is licensed under Apache 2.0.

Original post →

More from Multimodal

Multimodal channel →