Skip to content
master
Switch branches/tags
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

SampleRNN

A Tensorflow implementation of SampleRNN: An Unconditional End-to-End Neural Audio Generation Model.

Requirements

  • Tensroflow 1.0
  • Python 2.7
  • Librosa  

Dataset  

We used the pinao music of 74 minutes as the training corpus, and you can use any corpus containing ".wav" files to instead as well.
For Mandarin samples, we used human voice of 6 hours as the training corpus .

Samples

Pretrained model

FEATURES

  • 2-tier SampleRNN
  • 3-tier SampleRNN
  • Quantization in linear.
  • Quantization in mu-law.

Training

python train.py \
	--data_dir=./pinao-corpus \
	--silence_threshold=0.1 \
	--sample_size=102408 \
	--big_frame_size=8 \
	--frame_size=2 \
	--q_levels=256 \
	--rnn_type=GRU \
	--dim=1024 \
	--n_rnn=1 \
	--seq_len=520 \
	--emb_size=256 \
	--batch_size=64 \
	--optimizer=adam \
	--num_gpus=4

or

sh run.sh

Related projects

This work is based on the flowing implementations with some modifications:

About

Tensorflow implementation of SampleRNN

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •  

Languages