|
--- |
|
language: |
|
- en |
|
tags: |
|
- text-to-speech |
|
- tts |
|
- speech |
|
license: mit |
|
datasets: |
|
- Matthijs/cmu-arctic-xvectors |
|
--- |
|
|
|
# SpeechT5 TTS |
|
|
|
This is a re-upload of the Microsoft/SpeechT5_TTS model. |
|
|
|
## Model description |
|
|
|
SpeechT5 is a unified-modal speech and text model developed by Microsoft. This specific model is fine-tuned for text-to-speech tasks. |
|
|
|
## Usage |
|
|
|
```python |
|
from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan |
|
from datasets import load_dataset |
|
import torch |
|
import soundfile as sf |
|
|
|
processor = SpeechT5Processor.from_pretrained("YOUR_USERNAME/YOUR_REPO_NAME") |
|
model = SpeechT5ForTextToSpeech.from_pretrained("YOUR_USERNAME/YOUR_REPO_NAME") |
|
vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan") |
|
|
|
inputs = processor(text="Hello, how are you?", return_tensors="pt") |
|
|
|
embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation") |
|
speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0) |
|
|
|
speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder) |
|
|
|
sf.write("speech.wav", speech.numpy(), samplerate=16000) |
