CoquiTTS-Docker / app.py
Eren Gรถlge
Update app.py
3468586
raw
history blame
3.84 kB
import tempfile
from typing import Optional
from TTS.config import load_config
import gradio as gr
import numpy as np
from TTS.utils.manage import ModelManager
from TTS.utils.synthesizer import Synthesizer
MODELS = {}
SPEAKERS = {}
manager = ModelManager()
MODEL_NAMES = manager.list_tts_models()
# filter out multi-speaker models
filters = ["vctk", "your_tts"]
MODEL_NAMES = [model_name for model_name in MODEL_NAMES if not any(f in model_name for f in filters)]
# reorder models
MODEL_NAMES[0], MODEL_NAMES[1], MODEL_NAMES[2]= MODEL_NAMES[1], MODEL_NAMES[2], MODEL_NAMES[0]
print(MODEL_NAMES)
def tts(text: str, model_name: str, speaker_idx: str=None):
print(text, model_name)
# download model
model_path, config_path, model_item = manager.download_model(f"tts_models/{model_name}")
vocoder_name: Optional[str] = model_item["default_vocoder"]
# download vocoder
vocoder_path = None
vocoder_config_path = None
if vocoder_name is not None:
vocoder_path, vocoder_config_path, _ = manager.download_model(vocoder_name)
# init synthesizer
synthesizer = Synthesizer(
model_path, config_path, None, None, vocoder_path, vocoder_config_path,
)
# synthesize
if synthesizer is None:
raise NameError("model not found")
wavs = synthesizer.tts(text, speaker_idx)
# return output
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as fp:
synthesizer.save_wav(wavs, fp)
return fp.name
article= """
Visit us on Coqui.ai and drop a ๐ŸŒŸ to ๐Ÿ”—<a href="https://github.com/coqui-ai/TTS" target="_blank">CoquiTTS</a>.
<br/>
Run CoquiTTS locally for the best result. Check out our ๐Ÿ”—<a href="https://tts.readthedocs.io/en/latest/inference.html">documentation</a>.
```bash
$ pip install TTS
...
$ tts --list_models
...
$ tts --text "Text for TTS" --model_name "<type>/<language>/<dataset>/<model_name>" --out_path folder/to/save/output.wav
```
<img src="https://static.scarf.sh/a.png?x-pxid=1404a024-e647-4406-bb9a-4ade0c931182" />
<br/>
๐Ÿ‘‘ <b> Model contributors</b>
- <a href="https://github.com/nmstoker/" target="_blank">@nmstoker</a>
- <a href="https://github.com/kaiidams/" target="_blank">@kaiidams</a>
- <a href="https://github.com/WeberJulian/" target="_blank">@WeberJulian,</a>
- <a href="https://github.com/Edresson/" target="_blank">@Edresson</a>
- <a href="https://github.com/thorstenMueller/" target="_blank">@thorstenMueller</a>
- <a href="https://github.com/r-dh/" target="_blank">@r-dh</a>
- <a href="https://github.com/kirianguiller/" target="_blank">@kirianguiller</a>
- <a href="https://github.com/robinhad/" target="_blank">@robinhad</a>
- <a href="https://github.com/fkarabiber/" target="_blank">@fkarabiber</a>
- <a href="https://github.com/nicolalandro/" target="_blank">@nicolalandro</a>
๐Ÿ‘‰ Drop a โœจPRโœจ on ๐ŸธTTS to share a new model and have it included here.
"""
iface = gr.Interface(
fn=tts,
inputs=[
gr.inputs.Textbox(
label="Input Text",
default="Hello, how are you?",
),
gr.inputs.Radio(
label="Pick a TTS Model - (language/dataset/model_name)",
choices=MODEL_NAMES,
),
# gr.inputs.Dropdown(label="Select a speaker", choices=SPEAKERS, default=None)
# gr.inputs.Audio(source="microphone", label="Record your voice.", type="numpy", label=None, optional=False)
],
outputs=gr.outputs.Audio(label="Output"),
title="๐Ÿธ๐Ÿ’ฌ CoquiTTS Demo",
theme="grass",
description="๐Ÿธ๐Ÿ’ฌ Coqui TTS - a deep learning toolkit for Text-to-Speech, battle-tested in research and production.",
article=article,
allow_flagging=False,
flagging_options=['error', 'bad-quality', 'wrong-pronounciation'],
layout="vertical",
live=False
)
iface.launch(share=False)