Taras Sereda

taras.y.sereda@proton.me

I’m a Machine Learning Researcher, Engineer, and DIY Enthusiast. My career is built around Machine Learning research projects in vision and audio. At this moment, I’m providing consultancy services concerning speech and audio processing, particularly for building text-to-speech (TTS) systems. My interests lie in self-supervised methods and discrete representations of audio.

selected publications

paper

Pheme: Efficient and Conversational Speech Generation

Budzianowski, Paweł, Sereda, Taras, Cichy, Tomasz, and Vulić, Ivan

2024

Bib HTML PDF

@misc{budzianowski2024phemeefficientconversationalspeech,
  title = {Pheme: Efficient and Conversational Speech Generation},
  author = {Budzianowski, Paweł and Sereda, Taras and Cichy, Tomasz and Vulić, Ivan},
  year = {2024},
  eprint = {2401.02839},
  archiveprefix = {arXiv},
  primaryclass = {eess.AS},
}

paper

Transcribe, Align and Segment: Creating speech datasets for low-resource languages

Sereda, Taras

2024

Bib HTML PDF

@misc{sereda2024transcribealignsegmentcreating,
  title = {Transcribe, Align and Segment: Creating speech datasets for low-resource languages},
  author = {Sereda, Taras},
  year = {2024},
  eprint = {2406.12674},
  archiveprefix = {arXiv},
  primaryclass = {eess.AS},
  doi = {10.48550/arXiv.2406.12674},
}