gitgato commited on
Commit
cb4e037
·
verified ·
1 Parent(s): 11ed18a

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +6 -8
app.py CHANGED
@@ -3,15 +3,15 @@ import torch
3
  from datasets import load_dataset
4
  from transformers import pipeline, SpeechT5Processor, SpeechT5HifiGan, SpeechT5ForTextToSpeech
5
 
6
- model_id = "gitgato/tr-xtts" # update with your model id
7
- # pipe = pipeline("automatic-speech-recognition", model=model_id)
8
  model = SpeechT5ForTextToSpeech.from_pretrained(model_id)
9
  vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
10
  embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
11
  speaker_embeddings = torch.tensor(embeddings_dataset[7440]["xvector"]).unsqueeze(0)
12
 
13
  # checkpoint = "microsoft/speecht5_tts"
14
- processor = SpeechT5Processor.from_pretrained(model_id)
15
 
16
  replacements = [
17
  ("à", "a"),
@@ -30,7 +30,7 @@ replacements = [
30
 
31
  title = "Text-to-Speech"
32
  description = """
33
- Demo for text-to-speech translation in French. Demo uses [Sandiago21/speecht5_finetuned_facebook_voxpopuli_french](https://huggingface.co/Sandiago21/speecht5_finetuned_facebook_voxpopuli_french) checkpoint, which is based on Microsoft's
34
  [SpeechT5 TTS](https://huggingface.co/microsoft/speecht5_tts) model and is fine-tuned in French Audio dataset
35
  ![Text-to-Speech (TTS)"](https://geekflare.com/wp-content/uploads/2021/07/texttospeech-1200x385.png "Diagram of Text-to-Speech (TTS)")
36
  """
@@ -53,9 +53,7 @@ syntesize_speech_gradio = gr.Interface(
53
  synthesize_speech,
54
  inputs = gr.Textbox(label="Text", placeholder="Type something here..."),
55
  outputs=gr.Audio(),
56
- examples=["Hola, probando audio."],
57
  title=title,
58
  description=description,
59
- ).launch()
60
-
61
-
 
3
  from datasets import load_dataset
4
  from transformers import pipeline, SpeechT5Processor, SpeechT5HifiGan, SpeechT5ForTextToSpeech
5
 
6
+ model_id = "gitgato/tr-tts" # update with your model id
7
+ pipe = pipeline("text-to-speech", model="gitgato/tr-xtts")
8
  model = SpeechT5ForTextToSpeech.from_pretrained(model_id)
9
  vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
10
  embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
11
  speaker_embeddings = torch.tensor(embeddings_dataset[7440]["xvector"]).unsqueeze(0)
12
 
13
  # checkpoint = "microsoft/speecht5_tts"
14
+ processor = AutoProcessor.from_pretrained("gitgato/tr-xtts")
15
 
16
  replacements = [
17
  ("à", "a"),
 
30
 
31
  title = "Text-to-Speech"
32
  description = """
33
+ Demo for text-to-speech translation in French. Demo uses [gitgato/trtts](https://huggingface.co) checkpoint, which is based on Microsoft's
34
  [SpeechT5 TTS](https://huggingface.co/microsoft/speecht5_tts) model and is fine-tuned in French Audio dataset
35
  ![Text-to-Speech (TTS)"](https://geekflare.com/wp-content/uploads/2021/07/texttospeech-1200x385.png "Diagram of Text-to-Speech (TTS)")
36
  """
 
53
  synthesize_speech,
54
  inputs = gr.Textbox(label="Text", placeholder="Type something here..."),
55
  outputs=gr.Audio(),
56
+ examples=["Je n'entrerai pas dans les détails, mais je profiterai des secondes qui me restent pour exposer la position ALDE sur le marquage CE, un des points cruciaux de ce rapport."],
57
  title=title,
58
  description=description,
59
+ ).launch()