it seems something incompatible

thank you. would that be something I do different in the training? Or something I tweak in synthesis section?

the other models is working fine is just doesn't work the models from pipeline trainer notebook

oh gosh, so its not able to clone new models anymore? or am I misunderstanding? And if thats the case, are there any notebooks currently able to clone new voices?

you need to get the configuration from the training notebook (third-to-last cell, everything after

%%writefile blah.json

), save it to your computer as a text file, set the inferencing thing to custom config, and upload the text file when asked

because by default it doesn't expect a speaker embedding which the training notebook inserts

FileNotFound Error - was trying to make a voice out of Julie Berry. Even though I had list text file in correct folder, still had the error.

This is Tacotron 2 checkpoint 105 Does it needs more training or something wrong within the datasets or configuration?

transcription should be fine

nevermind

i should've checked mark the arpabet

You know what?

I never really knew the reason why active politicians can't be on Uberduck

slander

lately, i’ve done research on how ai-based tts tech can be used in the clinical setting, and i found a great study examining its use among a cohort of patients before and after a total laryngectomy the article doesn’t specify what software was used, but datasets were recorded from patients prior to surgery and used in the immediate postop period (before it is safe to try alternate methods of verbal communication such as esophageal speech, etc)

i thought i’d pass it along!

interesting!

thanks for sharing

is it possible to create a voice model out of just this audio clip?

one clip, no way

split into a few clips, more likely but still probably not

UnpicklingError Traceback (most recent call last) in () 5 print('cuDNN Benchmark:', hparams.cudnn_benchmark) 6 train(output_directory, log_directory, checkpoint_path, ----> 7 warm_start, n_gpus, rank, group_name, hparams, log_directory2) 3 frames /usr/local/lib/python3.7/dist-packages/torch/serialization.py in _legacy_load(f, map_location, pickle_module, **pickle_load_args) 775 "functionality.") 776 --> 777 magic_number = pickle_module.load(f, **pickle_load_args) 778 if magic_number != MAGIC_NUMBER: 779 raise RuntimeError("Invalid magic number; corrupt file?") UnpicklingError: invalid load key, '<'. Can someone help with this error from the tacotron training model?

please use the tacotron2 colab inside #841437191073955920 it should be the one at the very top

Okay, so I have the correct notebooks for Tacotron now but I am getting this error:

--------------------------------------------------------------------------- NameError Traceback (most recent call last) in () 23 try: ---> 24 initialized 25 except NameError: NameError: name 'initialized' is not defined During handling of the above exception, another exception occurred: Exception Traceback (most recent call last) 1 frames in () 143 return model, hparams 144 --> 145 model, hparams = get_Tactron2(TACOTRON2_ID) 146 previous_tt2_id = TACOTRON2_ID 147 in get_Tactron2(MODEL_ID) 129 gdown.download(d+MODEL_ID, tacotron2_pretrained_model, quiet=False) 130 if not exists(tacotron2_pretrained_model): --> 131 raise Exception("Tacotron2 model failed to download!") 132 # Load Tacotron2 and Config 133 hparams = create_hparams() Exception: Tacotron2 model failed to download!

Can anyone help with it?

this is great way of using TTS for people of disabilities I hope there'll be TTS with real time synthesizing feature