WHISPERSPEECH

Summary from the TLDR newsletter

WhisperSpeech is a text-to-speech system built by inverting Whisper. The models are currently only trained on the English LibreLight dataset, but future releases will target multiple languages. Samples of synthesized voices are available in the repository.