sulaimank commited on
Commit
4a68622
1 Parent(s): e981371

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +3 -39
README.md CHANGED
@@ -40,11 +40,11 @@ from speechbrain.inference.TTS import Tacotron2
40
  from speechbrain.inference.vocoders import HIFIGAN
41
 
42
  # Intialize TTS (tacotron2) and Vocoder (HiFIGAN)
43
- tacotron2 = Tacotron2.from_hparams(source="speechbrain/tts-tacotron2-ljspeech", savedir="tmpdir_tts")
44
- hifi_gan = HIFIGAN.from_hparams(source="speechbrain/tts-hifigan-ljspeech", savedir="tmpdir_vocoder")
45
 
46
  # Running the TTS
47
- mel_output, mel_length, alignment = tacotron2.encode_text("Mary had a little lamb")
48
 
49
  # Running Vocoder (spectrogram-to-waveform)
50
  waveforms = hifi_gan.decode_batch(mel_output)
@@ -53,42 +53,6 @@ waveforms = hifi_gan.decode_batch(mel_output)
53
  torchaudio.save('example_TTS.wav',waveforms.squeeze(1), 22050)
54
  ```
55
 
56
- If you want to generate multiple sentences in one-shot, you can do in this way:
57
-
58
- ```
59
- from speechbrain.pretrained import Tacotron2
60
- tacotron2 = Tacotron2.from_hparams(source="speechbrain/TTS_Tacotron2", savedir="tmpdir")
61
- items = [
62
- "A quick brown fox jumped over the lazy dog",
63
- "How much wood would a woodchuck chuck?",
64
- "Never odd or even"
65
- ]
66
- mel_outputs, mel_lengths, alignments = tacotron2.encode_batch(items)
67
-
68
- ```
69
-
70
- ### Inference on GPU
71
- To perform inference on the GPU, add `run_opts={"device":"cuda"}` when calling the `from_hparams` method.
72
-
73
- ### Training
74
- The model was trained with SpeechBrain.
75
- To train it from scratch follow these steps:
76
- 1. Clone SpeechBrain:
77
- ```bash
78
- git clone https://github.com/speechbrain/speechbrain/
79
- ```
80
- 2. Install it:
81
- ```bash
82
- cd speechbrain
83
- pip install -r requirements.txt
84
- pip install -e .
85
- ```
86
- 3. Run Training:
87
- ```bash
88
- cd recipes/LJSpeech/TTS/tacotron2/
89
- python train.py --device=cuda:0 --max_grad_norm=1.0 --data_folder=/your_folder/LJSpeech-1.1 hparams/train.yaml
90
- ```
91
- You can find our training results (models, logs, etc) [here](https://drive.google.com/drive/folders/1PKju-_Nal3DQqd-n0PsaHK-bVIOlbf26?usp=sharing).
92
 
93
  ### Limitations
94
  The SpeechBrain team does not provide any warranty on the performance achieved by this model when used on other datasets.
 
40
  from speechbrain.inference.vocoders import HIFIGAN
41
 
42
  # Intialize TTS (tacotron2) and Vocoder (HiFIGAN)
43
+ tacotron2 = Tacotron2.from_hparams(source="Sulaimank/tts-tacotron2-commonvoice-single-female", savedir="tmpdir_tts")
44
+ hifi_gan = HIFIGAN.from_hparams(source="Sulaimank/tts-hifigan-commonvoice-single-female", savedir="tmpdir_vocoder")
45
 
46
  # Running the TTS
47
+ mel_output, mel_length, alignment = tacotron2.encode_text("Obwedda ndowooza wagenze.")
48
 
49
  # Running Vocoder (spectrogram-to-waveform)
50
  waveforms = hifi_gan.decode_batch(mel_output)
 
53
  torchaudio.save('example_TTS.wav',waveforms.squeeze(1), 22050)
54
  ```
55
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
56
 
57
  ### Limitations
58
  The SpeechBrain team does not provide any warranty on the performance achieved by this model when used on other datasets.