app.py

app.py

@@ -1,14 +1,77 @@
+from typing import Optional
+
+from TTS.utils.manage import ModelManager
+from TTS.utils.synthesizer import Synthesizer
 import gradio as gr
+import tempfile
+
+
+
+MODEL_NAMES = [
+    "en/ek1/tacotron2",
+    "en/ljspeech/tacotron2-DDC",
+    # "en/ljspeech/tacotron2-DDC_ph",
+    # "en/ljspeech/glow-tts",
+    # "en/ljspeech/tacotron2-DCA",
+    # "en/ljspeech/speedy-speech-wn",
+    # "en/ljspeech/vits",
+    # "en/vctk/sc-glow-tts",
+    # "en/vctk/vits",
+    # "en/sam/tacotron-DDC",
+    # "es/mai/tacotron2-DDC",
+    "fr/mai/tacotron2-DDC",
+    "zh-CN/baker/tacotron2-DDC-GST",
+    "nl/mai/tacotron2-DDC",
+    "de/thorsten/tacotron2-DCA",
+    # "ja/kokoro/tacotron2-DDC",
+]
+MODELS = {}
+
+manager = ModelManager()
+
+for MODEL_NAME in MODEL_NAMES:
+    print(f"downloading {MODEL_NAME}")
+    model_path, config_path, model_item = manager.download_model(f"tts_models/{MODEL_NAME}")
+    vocoder_name: Optional[str] = model_item["default_vocoder"]
+    vocoder_path = None
+    vocoder_config_path = None
+    if vocoder_name is not None:
+        vocoder_path, vocoder_config_path, _ = manager.download_model(vocoder_name)
+
+    synthesizer = Synthesizer(
+        model_path, config_path, None, vocoder_path, vocoder_config_path,
+    )
+    MODELS[MODEL_NAME] = synthesizer
+
+
+def tts(text: str, model_name: str):
+    print(text, model_name)
+    synthesizer = MODELS.get(model_name, None)
+    if synthesizer is None:
+        raise NameError("model not found")
+    wavs = synthesizer.tts(text)
+    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as fp:
+        synthesizer.save_wav(wavs, fp)
+        return fp.name
+
 
-def greet(name):
-    print(gr.__version__)
-    return "Hello " + gr.__version__ + "!!"
 
 iface = gr.Interface(
-    fn=greet,
-    inputs="text",
-    outputs="text",
-    verbose=True,
-    # enable_queue=True,
+    fn=tts,
+    inputs=[
+        gr.inputs.Textbox(
+            label="Input",
+            default="Hello, how are you?",
+        ),
+        gr.inputs.Radio(
+            label="Pick a TTS Model",
+            choices=MODEL_NAMES,
+        ),
+    ],
+    outputs=gr.outputs.Audio(label="Output"),
+    title="🐸💬 - Coqui TTS",
+    theme="huggingface",
+    description="🐸💬 - a deep learning toolkit for Text-to-Speech, battle-tested in research and production",
+    article="more info at https://github.com/coqui-ai/TTS",
 )
 iface.launch()

requirements.txt

@@ -0,0 +1 @@
+TTS