Spaces:
Running
Running
File size: 3,628 Bytes
f23c138 cea6632 c8c05d4 49696ae cea6632 ed7f208 cea6632 49696ae cea6632 49696ae cea6632 1d606bc cea6632 49696ae cea6632 404e64b 1d606bc 404e64b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 |
import os
import gradio as gr
import numpy as np
import torch
from InferenceInterfaces.Meta_FastSpeech2 import Meta_FastSpeech2
os.system("pip uninstall -y gradio")
os.system("pip install gradio==2.7.5.2")
def float2pcm(sig, dtype='int16'):
"""
https://gist.github.com/HudsonHuang/fbdf8e9af7993fe2a91620d3fb86a182
"""
sig = np.asarray(sig)
if sig.dtype.kind != 'f':
raise TypeError("'sig' must be a float array")
dtype = np.dtype(dtype)
if dtype.kind not in 'iu':
raise TypeError("'dtype' must be an integer type")
i = np.iinfo(dtype)
abs_max = 2 ** (i.bits - 1)
offset = i.min + abs_max
return (sig * abs_max + offset).clip(i.min, i.max).astype(dtype)
class TTS_Interface:
def __init__(self):
self.device = "cuda" if torch.cuda.is_available() else "cpu"
self.model = Meta_FastSpeech2(device=self.device)
def read(self, prompt, language, path_to_audio):
language_id_lookup = {
"English" : "en",
"German" : "de",
"Greek" : "el",
"Spanish" : "es",
"Finnish" : "fi",
"Russian" : "ru",
"Hungarian": "hu",
"Dutch" : "nl",
"French" : "fr"
}
self.model.set_language(language_id_lookup[language])
if path_to_audio is not None:
try:
self.model.set_utterance_embedding(path_to_audio)
except RuntimeError:
pass
wav = self.model(prompt)
return 48000, float2pcm(wav.cpu().numpy())
meta_model = TTS_Interface()
article = "<p style='text-align: left'>This is still a work in progress, models will be exchanged for better ones as soon as they are done. All of those languages are spoken by a single model. Speakers can be transferred across languages. More languages will be added soon.</p><p style='text-align: center'><a href='https://github.com/DigitalPhonetics/IMS-Toucan' target='_blank'>Click here to learn more about the IMS Toucan Speech Synthesis Toolkit</a></p>"
iface = gr.Interface(fn=meta_model.read,
inputs=[gr.inputs.Textbox(lines=2, placeholder="write what you want the synthesis to read here...", label=" "),
gr.inputs.Dropdown(['English',
'German',
'Greek',
'Spanish',
'Finnish',
'Russian',
'Hungarian',
'Dutch',
'French'], type="value", default='English', label="Language Selection"),
gr.inputs.Audio(source="microphone",
optional=True,
label="Make the TTS imitate your Voice (optional, press once to start recording and again to stop)",
type="filepath")],
outputs=gr.outputs.Audio(type="numpy", label=None),
layout="vertical",
title="IMS Toucan Multilingual Multispeaker Demo",
thumbnail="Utility/toucan.png",
theme="default",
allow_flagging="never",
allow_screenshot=False,
article=article)
iface.launch(enable_queue=True)
|