| import audiofile | |
| import numpy as np | |
| from audiocraft import AudioGen | |
| text_list = ['dogs barging in the street', | |
| 'cats meowing'] | |
| sound_generator = AudioGen(duration=0.24, | |
| device='cuda:0').to('cuda:0').eval() | |
| x = sound_generator.generate(text_list) # [bs, 1, 7680] | |
| # print('demo', x.shape) | |
| x = x[1, :, :].detach().cpu().numpy() | |
| # x /= np.abs(x).max() + 1e-7 # inside generate() | |
| audiofile.write('del_seane.wav', x, 16000) | |