Update app.py

app.py

@@ -1,45 +1,107 @@
 import gradio as gr
 import time
+import urllib.request
+from pathlib import Path
+import os
 import torch
 import scipy.io.wavfile
 from espnet2.bin.tts_inference import Text2Speech
 from espnet2.utils.types import str_or_none
 
-tagen = "https://huggingface.co/wietsedv/tacotron2-gronings/resolve/main/tts_ljspeech_finetune_tacotron2.v5_train.loss.ave.zip" 
-vocoder_tagen = "parallel_wavegan/ljspeech_parallel_wavegan.v3" 
-
-text2speechen = Text2Speech.from_pretrained(
-    model_tag=tagen,
-    vocoder_tag=vocoder_tagen,
-    device="cpu",
-    threshold=0.5,
-    minlenratio=0.0,
-    maxlenratio=10.0,
-    use_att_constraint=True,
-    backward_window=1,
-    forward_window=4,
+
+# def load_model(model_tag, vocoder_tag):
+#     from espnet_model_zoo.downloader import ModelDownloader
+
+#     kwargs = {}
+
+#     # Model
+#     d = ModelDownloader()
+#     kwargs = d.download_and_unpack(model_tag)
+
+#     # Vocoder
+#     download_dir = Path(os.path.expanduser("~/.cache/parallel_wavegan"))
+#     vocoder_dir = download_dir / vocoder_tag
+#     os.makedirs(vocoder_dir, exist_ok=True)
+
+#     kwargs["vocoder_config"] = vocoder_dir / "config.yml"
+#     if not kwargs["vocoder_config"].exists():
+#         urllib.request.urlretrieve(f"https://huggingface.co/{vocoder_tag}/resolve/main/config.yml", kwargs["vocoder_config"])
+
+#     kwargs["vocoder_file"] = vocoder_dir / "checkpoint-50000steps.pkl"
+#     if not kwargs["vocoder_file"].exists():
+#         urllib.request.urlretrieve(f"https://huggingface.co/{vocoder_tag}/resolve/main/checkpoint-50000steps.pkl", kwargs["vocoder_file"])
+
+#     return Text2Speech(
+#         **kwargs,
+#         device="cpu",
+#         threshold=0.5,
+#         minlenratio=0.0,
+#         maxlenratio=10.0,
+#         use_att_constraint=True,
+#         backward_window=1,
+#         forward_window=4,
+#     )
+
+# gos_text2speech = load_model('https://huggingface.co/wietsedv/tacotron2-gronings/resolve/main/tts_ljspeech_finetune_tacotron2.v5_train.loss.ave.zip', 'wietsedv/parallelwavegan-gronings')
+# nld_text2speech = load_model('https://huggingface.co/wietsedv/tacotron2-dutch/resolve/main/tts_ljspeech_finetune_tacotron2.v5_train.loss.ave.zip', 'wietsedv/parallelwavegan-dutch')
+
+gos_text2speech = Text2Speech.from_pretrained(
+   model_tag="https://huggingface.co/wietsedv/tacotron2-gronings/resolve/main/tts_ljspeech_finetune_tacotron2.v5_train.loss.ave.zip",
+   vocoder_tag="parallel_wavegan/ljspeech_parallel_wavegan.v3",
+   device="cpu",
+   threshold=0.5,
+   minlenratio=0.0,
+   maxlenratio=10.0,
+   use_att_constraint=True,
+   backward_window=1,
+   forward_window=4,
 )
+nld_text2speech = Text2Speech.from_pretrained(
+   model_tag="https://huggingface.co/wietsedv/tacotron2-dutch/resolve/main/tts_ljspeech_finetune_tacotron2.v5_train.loss.ave.zip",
+   vocoder_tag="parallel_wavegan/ljspeech_parallel_wavegan.v3",
+   device="cpu",
+   threshold=0.5,
+   minlenratio=0.0,
+   maxlenratio=10.0,
+   use_att_constraint=True,
+   backward_window=1,
+   forward_window=4,
+)
+#eng_text2speech = Text2Speech.from_pretrained(
+#    model_tag="kan-bayashi/ljspeech_tacotron2",
+#    vocoder_tag="parallel_wavegan/ljspeech_parallel_wavegan.v3",
+#    device="cpu",
+#    threshold=0.5,
+#    minlenratio=0.0,
+#    maxlenratio=10.0,
+#    use_att_constraint=True,
+#    backward_window=1,
+#    forward_window=4,
+#)
 
 def inference(text,lang):
   with torch.no_grad():
       if lang == "gronings":
-          wav = text2speechen(text)["wav"]
-          scipy.io.wavfile.write("out.wav",text2speechen.fs , wav.view(-1).cpu().numpy())
-         
-  return  "out.wav"
-title = "ESPnet2-TTS"
-description = "Gradio demo for ESPnet2-TTS: Extending the Edge of TTS Research. To use it, simply add your audio, or click one of the examples to load them. Read more at the links below."
-article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2110.07840' target='_blank'>ESPnet2-TTS: Extending the Edge of TTS Research</a> | <a href='https://github.com/espnet/espnet' target='_blank'>Github Repo</a></p>"
+          wav = gos_text2speech(text)["wav"]
+          scipy.io.wavfile.write("out.wav", gos_text2speech.fs , wav.view(-1).cpu().numpy())
+      if lang == "dutch":
+          wav = nld_text2speech(text)["wav"]
+          scipy.io.wavfile.write("out.wav", nld_text2speech.fs , wav.view(-1).cpu().numpy())
+      #if lang == "english":
+      #    wav = eng_text2speech(text)["wav"]
+      #    scipy.io.wavfile.write("out.wav", eng_text2speech.fs , wav.view(-1).cpu().numpy())
+
+  return  "out.wav", "out.wav"
 
-examples=[['This paper describes ESPnet2-TTS, an end-to-end text-to-speech (E2E-TTS) toolkit. ESPnet2-TTS extends our earlier version, ESPnet-TTS, by adding many new features, including: on-the-fly flexible pre-processing, joint training with neural vocoders, and state-of-the-art TTS models with extensions like full-band E2E text-to-waveform modeling, which simplify the training pipeline and further enhance TTS performance. The unified design of our recipes enables users to quickly reproduce state-of-the-art E2E-TTS results',"english"]]
+title = "GroTTS"
+examples = [
+  ['Ze gingen mit klas noar waddendiek, over en deur bragel lopen.', 'gronings']
+]
 
 gr.Interface(
-    inference, 
-    [gr.inputs.Textbox(label="input text",lines=10),gr.inputs.Radio(choices=["gronings"], type="value", default="gronings", label="language")], 
-    gr.outputs.Audio(type="file", label="Output"),
+    inference,
+    [gr.inputs.Textbox(label="input text", lines=3), gr.inputs.Radio(choices=["gronings", "dutch"], type="value", default="gronings", label="language")], 
+    [gr.outputs.Audio(type="file", label="Output"), gr.outputs.File()],
     title=title,
-    description=description,
-    article=article,
-    enable_queue=True,
     examples=examples
-    ).launch(debug=True)
+    ).launch(enable_queue=True)