espnet2.sds.tts.espnet_tts.ESPnetTTSModel

source

class espnet2.sds.tts.espnet_tts.ESPnetTTSModel(tag: str = 'kan-bayashi/ljspeech_vits', device: str = 'cuda')

Bases: AbsTTS

ESPnet TTS.

A class to initialize and manage a ESPnet’s pre-trained text-to-speech (TTS) model.

This class:

1. Downloads and sets up a pre-trained TTS model using the ESPnet Model Zoo.
2. Supports various TTS configurations, including multi speaker TTS using speaker embeddings and speaker IDs.

• Parameters:
  • tag (str , optional) – The model tag for the pre-trained TTS model. Defaults to “kan-bayashi/ljspeech_vits”.
  • device (str , optional) – The computation device for running inference. Defaults to “cuda”.
• Raises:ImportError – If the required espnet_model_zoo library is not installed.

forward(transcript: str) → Tuple[int, ndarray]

Converts a text transcript into an audio waveform using a pre-trained ESPnet-TTS model.

• Parameters:transcript (str) – The input text to be converted into speech.
• Returns: A tuple containing:
  • The sample rate of the audio (int).
  • The generated audio waveform as a NumPy array of type int16.
• Return type: Tuple[int, np.ndarray]

warmup()

Perform a single forward pass with dummy input to pre-load and warm up the model.