Empero Research releases PiVo-1.4M, a 1.4M-parameter text-to-speech model for local use
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.
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
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.
empero-ai/PiVo-1.4M · Hugging Face
huggingface.co · 2 November 2026
Loading the full article…
This text was published by huggingface.co. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Reddit discussionreddit.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Researchers prune 729 attention heads to cut GPT-2-XL’s memorisation by 91.5% · 1 src
- arXiv paper finds shortcuts distort RAG benchmarking accuracy · 1 src
- Researchers propose adaptive AI personalization for shifting user preferences · 1 src
- Researchers introduce CoFiBreak to improve jailbreak attack diversity in LLMs · 1 src
- Researchers release Qistas protocol for testing LLM fairness across cultures · 1 src
Comments
via GitHub Discussions