Create text interface for TTS

[jerivl]

Issue: Need to be able to run TTS easily

Task: Create function that creates audio from text using pytorch implementation of tacotron2 https://github.com/NVIDIA/tacotron2

Desired input: Input string, sample rate

Desired output: Numpy array containing audio data

Prefer sample rate = 44.1k but 16k at minimum

[hamzabhatti12]

Here is the traceback from the error I am getting:

Using cache found in C:\Users\Hamza/.cache\torch\hub\nvidia_DeepLearningExamples_torchhub Traceback (most recent call last): File "C:\Users\Hamza\Documents\GitHub\Deepcut\tests\tacotron\tacotron.py", line 3, in tacotron2 = torch.hub.load('nvidia/DeepLearningExamples:torchhub', 'nvidia_tacotron2') File "F:\ProgramData\Anaconda3\lib\site-packages\torch\hub.py", line 353, in load model = entry(*args, **kwargs) File "C:\Users\Hamza/.cache\torch\hub\nvidia_DeepLearningExamples_torchhub/hubconf.py", line 131, in nvidia_tacotron2 from PyTorch.SpeechSynthesis.Tacotron2.models import lstmcell_to_float, batchnorm_to_float File "C:\Users\Hamza/.cache\torch\hub\nvidia_DeepLearningExamples_torchhub\PyTorch\SpeechSynthesis\Tacotron2\models.py", line 32, in from tacotron2.model import Tacotron2 File "C:\Users\Hamza\Documents\GitHub\Deepcut\tests\tacotron\tacotron2\model.py", line 6, in from layers import ConvNorm, LinearNorm ModuleNotFoundError: No module named 'layers'

[jerivl]

Have you tried pasting in the code from https://github.com/NVIDIA/tacotron2/blob/master/model.py into your working directory?

It just seems like a linking problem to me, but I can't remember if you said that there was something else in our meetings.