Streaming Inference with Convolutional Layers
In this post, we explore how to apply convolutional layers to infinitely long inputs, specifically focusing on how to process inputs in chunks to minimize la...
Posts
2024
2023
Dissecting BARK
Things started to get stale after the ubiquitous switch to Neural Text-to-Speech. A long-awaited leap forward was introduced thanks to ideas from the blossom...
Zero-shot speech generation benchmark
Synthesizing speech with a speaker identity not seen during training presents a significant challenge. Traditionally, achieving this required extensive train...
Українська мова в Balacoon
Швидкий, зручний та якісний нейромережевий синтез українського мовлення тепер в Balacoon. Інтеграція бібліотеки синтезу ще ніколи не була такою простою: Pyth...
Balacoon TTS on-device
Neural text-to-speech brought unprecedented improvements in the naturalness of synthetic speech. But it came with a cost. While parametric and concatenative ...
Balacoon TTS as a service
In recent years, text-to-speech technology has made tremendous strides, thanks in large part to advances in machine learning and artificial intelligence. As ...
Balacoon TTS version 0.1.0
We’re excited to announce the release of Balacoon TTS 0.1.0, the latest version of our text-to-speech package. This new version includes two major updates th...
Balacoon phonemeset
Text-to-speech assumes the implicit or explicit conversion of input text into a sequence of sounds to be pronounced. Defining a set of all possible sounds (o...
en-US abbreviation detection
Detecting abbreviations is crucial for proper text normalization and subsequent pronunciation generation. In broad terms, “abbreviation” means shortening, co...