Model reference · open weights

neutts-nano

neutts-nano is an open-weight audio or speech model from neuphonic. neutts-nano (FP32) weighs 457 MB; the smallest configuration that runs it is RTX 3060 12 GB.

• neutts-nano is a 229M parameter text-to-speech model developed by neuphonic for on-device voice synthesis.
• It supports English and features instant voice cloning from short audio references, optimized for real-time generation on CPUs and edge devices.
• The model uses a compact language model backbone with the NeuCodec audio codec and is available in GGUF format under an other license.

Summary of the neuphonic/neutts-nano model card, 2026-10-01

Hugging Face ↗