Zero-shot Text-to-Speech: How Does the Performance of HierSpeech++ Fare With Other Baselines? | HackerNoon
Briefly

from Hackernoon8 months ago
In our evaluation, HierSpeech++ consistently exhibited superior performance across various subjective and objective metrics for zero-shot text-to-speech synthesis, surpassing even ground truth in terms of naturalness.
Hackernoonhttps://hackernoon.com/zero-shot-text-to-speech-how-does-the-performance-of-hierspeech-fare-with-other-baselines
While HierSpeech++ showed remarkable results in naturalness and overall performance, XTTS held an edge in pMOS, indicating that further refinement in our model could enhance future outcomes.
Hackernoonhttps://hackernoon.com/zero-shot-text-to-speech-how-does-the-performance-of-hierspeech-fare-with-other-baselines
Read at Hackernoon
#text-to-speech #zero-shot-learning #hierarchical-models #speech-synthesis #voice-conversion
[
|
]