Update app.py

app.py

@@ -9,41 +9,64 @@ from espnet2.bin.tts_inference import Text2Speech
 from espnet2.utils.types import str_or_none
 
 
-def load_model(model_tag, vocoder_tag):
-    from espnet_model_zoo.downloader import ModelDownloader
+# def load_model(model_tag, vocoder_tag):
+#     from espnet_model_zoo.downloader import ModelDownloader
 
-    kwargs = {}
+#     kwargs = {}
 
-    # Model
-    d = ModelDownloader()
-    kwargs = d.download_and_unpack(model_tag)
+#     # 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)
+#     # 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_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"])
+#     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,
-    )
+#     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 = 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",