Voice
training dataset voices and custom mix
These are the audiobook readers the model was trained on. Their voices are theirs: see the licence note below before using a recognisable reader. Moving a slider makes a custom mix.
advanced: pitch range, temperature, tempo, ODE steps, backend
temperature
tempo (length scale)
pronunciation dictionary
One rule per line:
word -> how to say it in Polish spelling. Whole words, case-insensitive, applied before phonemization. Saved in this browser.loading…
Licence and voice rights. Model weights: CC BY-SA 4.0, trained on Wolne Lektury audiobooks (CC BY-SA 3.0 PL) and AZON (CC BY-SA 4.0); every book and reader is credited in the model's ATTRIBUTION.md. The licence covers the recordings and the weights, not the voices: a voice is a personal attribute of the reader, and you have no right to use a recognisable voice of a real person (for example a single reader) commercially, to impersonate them or to mislead listeners without their consent. Playground code MIT; espeak-ng GPL-3.0.
Your own voice. The published checkpoint can be fine-tuned on one to two hours of clean recordings of one speaker (about an hour on a single GPU), followed by a short vocoder fine-tune; use recordings of your own voice or of someone who has agreed. Step-by-step: RECIPE.md, scripts in tts-pl-playground.
History (newest first). Each clip keeps its settings; send it back to the effects panel with “→ effects”.
Effects
Effects apply to the clip playing in History (the newest one until you play another). Its ⬇ wav saves what you hear.