Oriel — AI & technology worth noticing, every day.

ConwayResearch/Underdog-Woof

AI model · https://huggingface.co/ConwayResearch/Underdog-Woof

First seen by Oriel on .

AI summary

Underdog Woof is a 40.6 MB, two-voice English text-to-speech model that runs locally on a CPU and reports 1.42% word error rate on Gradium TTS Hard Cases.

  • The model includes Juno and Mira, produces 24 kHz mono audio, and uses a NumPy runtime without a GPU, server, or API key.
  • A built-in normalizer handles numbers, dates, times, money, codes, symbols, and selected names; long text is split into shorter pieces and problematic audio can be regenerated.
  • On one Apple M5 Max CPU thread, Juno runs at 13.8x real time with 402 MB peak memory, while Mira runs at 7.2x with 601 MB.
  • The reported evaluation covers 500 English sentences; Woof's Mira voice scored 1.42% word error rate versus 1.89% for Fish Audio Drama 3.
  • Use is free for non-commercial purposes with attribution; commercial use requires a license, and using Woof or its audio to train another speech or audio model is prohibited.

#text-to-speech #numpy #speech synthesis

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Likes

Total likes reported by the source.

  • Hugging Face Models: 268, latest observation ; +166 over observed period since

Downloads (30d)

Rolling 30-day downloads. A change in this value is not new downloads during the observed period.

  • Hugging Face Models: 10, latest observation ; 0 over observed period since

Aliases

SourceExternal IDURL
Hugging Face ModelsConwayResearch/Underdog-Woofhttps://huggingface.co/ConwayResearch/Underdog-Woof Read discussion