Speech Synthesis TRIZ Cases
- Text-to-Speech Synthesis with Asynchronous Prosodic Phrase Processing
A thread pool processes prosodic phrases asynchronously, enabling earlier audio playback while maintaining speech synthesis quality.
- Speech Generation With Feature Fusion for Low-Data Voice Cloning
A speech model combines speaker, language, and text features to simplify training for low-data TTS and voice conversion.
- Audio Channel Augmentation with Impulse Responses for Low-Resource Speech
An audio corpus yields channel impulse responses that lower augmentation cost and improve recognition for noisy, low-resource recordings.
- Encoder and Decoder Disentangling Style and Content for Low-Bitrate Speech
Separate encoders isolate speaker identity from phonetic content for efficient, private speech reconstruction at low bitrates.
- Vocoder Subframe Prediction for Faster Audio Synthesis
Frequency subframes and synchronous adjacent-sample prediction reduce vocoder loops for faster, more efficient audio synthesis.
- Speech Synthesis Decoder for User-Controlled Speech Style Variation
A supervised decoder combines text with user voice features to generate diverse speech styles from the same learning data.