search

r9y9 / nnmnkwii_gallery

A collection of examples demonstrating how we can build speech synthesis systems using nnmnkwii.
https://github.com/r9y9/nnmnkwii
Other
71 stars 21 forks source link

Feature extraction problem #1

Closed nd39 closed 5 years ago

nd39 commented 6 years ago

Hi, I was trying your code on my own database, however an error came out in the stage of feature extraction when using the commande line ! python ./scripts/prepare_features.py ./data/slt_arctic_full_data which is :

Traceback (most recent call last): File "./scripts/prepare_features.py", line 226, in print("Acoustic linguistic feature dim", X_acoustic[0].shape) File "/home/ubuntu/anaconda3/lib/python3.6/site-packages/nnmnkwii/datasets/init.py", line 146, in getitem return self.__collect_features(paths) File "/home/ubuntu/anaconda3/lib/python3.6/site-packages/nnmnkwii/datasets/init.py", line 131, in __collect_features return self.file_data_source.collect_features(paths) File "./scripts/prepare_features.py", line 75, in collect_features subphone_features=self.subphone_features) File "/home/ubuntu/anaconda3/lib/python3.6/site-packages/nnmnkwii/frontend/merlin.py", line 511, in linguistic_features return load_labels_with_phone_alignment(hts_labels, args, **kwargs) File "/home/ubuntu/anaconda3/lib/python3.6/site-packages/nnmnkwii/frontend/merlin.py", line 238, in load_labels_with_phone_alignment subphone_features, add_frame_features)) ValueError: Combination of subphone_features and add_frame_features is not supported: full, True

Coud you help me solve it. Thank you in advance.

r9y9 commented 6 years ago

Which notebook did you run?

nd39 commented 6 years ago

the notebook of Bidirectional-LSTM based RNNs for text-to-speech synthesis

r9y9 commented 6 years ago

Sorry for the delay. I tried but cannot reproduce. What changes did you make? What does your data look like?

nd39 commented 6 years ago

thank you, I solve it, the problem was my label state alignment. I have another question, if i want to use the label phone alignment insteed of label state alignment, what changes should i make.

r9y9 commented 6 years ago

http://nbviewer.jupyter.org/github/r9y9/nnmnkwii_gallery/blob/master/notebooks/tts/jp-01-DNN-based%20statistical%20speech%20synthesis.ipynb uses phone-level alignment for Japanese TTS and it might help to see what changes you need.

nd39 commented 6 years ago

Thank you for helping me.

mrgloom commented 5 years ago

Get same error using ./nnmnkwii_gallery/data/slt_arctic_full_data/label_phone_align/arctic_b0538.lab but for ./nnmnkwii_gallery/data/slt_arctic_full_data/label_state_align/arctic_b0538.lab it's ok.

stale[bot] commented 5 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.