Single-speaker Hebrew/English speech dataset.
The voice was created with Qwen Voice Design. The English split uses LJSpeech text synthesized with Qwen3-TTS. The Hebrew split was created with Chatterbox, transcribed into text and IPA, and filtered to keep rows where both ASR passes agreed and the text/IPA alignment passed an FST-style check.
Audio files are in wav/.
metadata.csv has no header and uses | as the separator:… See the full description on the dataset page:
https://huggingface.co/datasets/thewh1teagle/synthetic-multilingual-speech.
