Dataset · machinekind
Polish conversational synthetic speech (TTS)
machinekind/polish-synthetic-speech-voxcpm2-v2
Nearly 23 hours of conversational Polish audio across two voices for distilling compact speech synthesis models.
What's inside
The dataset contains 19,646 audio clips totaling 22.8 hours in Polish with transcriptions, voiced by two synthetic voices using the VoxCPM2 model. The recordings simulate natural humanoid robot speech (reactions, questions, short stories) and have undergone individual quality checks.
What you can do with it
- Distilling large TTS models into compact versions for on-device deployment
- Training and fine-tuning conversational Polish speech synthesis models (e.g., Matcha-TTS)
- Tasks
- Text → speech
- Languages
- Polish
- Size
- 10K–100K rows
- Downloads this month
- 597
- Likes
- 4
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("machinekind/polish-synthetic-speech-voxcpm2-v2")