# Empero Research releases PiVo-1.4M, a 1.4M-parameter text-to-speech model for local use

Digest AI · Research · published 2026-11-02T15:42:27Z

Canonical: https://digestai.news/story/empero-research-releases-pivo-1-4m-a-1-4m-parameter-text-to-speech-mod

## Summary

Empero Research released PiVo-1.4M, a lightweight text-to-speech (TTS) model with **1,401,562 parameters** designed for local CPU inference. The model includes two synthetic English voices, Pi and Vo, and runs entirely offline using NumPy and ONNX Runtime. It avoids PyTorch dependencies and does not require network access during inference, making it suitable for privacy-sensitive applications.

PiVo achieves **intelligible speech** but retains a metallic or robotic texture in some sentences. The model processes input text via a CMU pronouncing dictionary and supports custom pronunciations via ARPAbet overrides. Input is limited to **2,000 characters** and **512 frontend tokens**, with a maximum utterance length of **30 seconds**. The model’s acoustic design is inspired by Neuroblast, and its vocoder is derived from FARGAN. Training data for Pi and Vo was generated using Qwen3-TTS, resulting in **8,628 utterances** (~12.29 hours of audio). The model’s weights are released under **Apache-2.0**, while the vocoder and pronunciation dictionary use separate licenses.

## Key points

- PiVo-1.4M runs locally on CPU with **1.4M parameters**, no network required, and supports two synthetic English voices
- Model achieves **intelligible speech** but retains robotic texture; input limited to **2,000 chars/512 tokens per utterance**
- Released under **Apache-2.0**, with training data and teacher models excluded; supports custom pronunciation via ARPAbet

## Why it matters

A lightweight, offline TTS model with synthetic voices could enable privacy-focused applications without cloud dependencies, though intelligibility trade-offs remain.

## Sources

1. [empero-ai/PiVo-1.4M · Hugging Face](https://huggingface.co/empero-ai/PiVo-1.4M) (huggingface.co, 2026-11-02, primary source)

## Cite

Digest AI, "Empero Research releases PiVo-1.4M, a 1.4M-parameter text-to-speech model for local use", 2 November 2026, https://digestai.news/story/empero-research-releases-pivo-1-4m-a-1-4m-parameter-text-to-speech-mod

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/empero-research-releases-pivo-1-4m-a-1-4m-parameter-text-to-speech-mod.json