|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
import gradio as gr |
|
import subprocess |
|
import os |
|
import json |
|
import uuid |
|
import requests |
|
from pypipertts import PyPiper |
|
pp=PyPiper() |
|
|
|
my_examples = """ |
|
An der Zäit hunn sech den Nordwand an d’Sonn gestridden, wie vun hinnen zwee wuel méi staark wier, |
|
wéi e Wanderer, deen an ee waarme Mantel agepak war, iwwert de Wee koum. |
|
Si goufen sech eens, datt deejéinege fir dee Stäerkste gëlle sollt, deen de Wanderer forcéiere géif, säi Mantel auszedoen. |
|
Den Nordwand huet mat aller Force geblosen, awer wat e méi geblosen huet, wat de Wanderer sech méi a säi Mantel agewéckelt huet. |
|
Um Enn huet den Nordwand säi Kampf opginn. Dunn huet d’Sonn d’Loft mat hire frëndleche Strale gewiermt, |
|
a schonn no kuerzer Zäit huet de Wanderer säi Mantel ausgedoen. Do huet den Nordwand missen zouginn, |
|
datt d’Sonn vun hinnen zwee dee Stäerkste wier.""" |
|
|
|
speeds = [ |
|
"ganz lues", |
|
"lues", |
|
"normal", |
|
"schnell", |
|
"ganz schnell" |
|
] |
|
|
|
stops = [ |
|
"ganz kuerz", |
|
"kuerz", |
|
"mëttel", |
|
"laang", |
|
"ganz laang" |
|
] |
|
|
|
def change_speed(choice): |
|
|
|
if choice=="ganz lues": |
|
speed=1.8 |
|
elif choice=="lues": |
|
speed=1.4 |
|
elif choice=="normal": |
|
speed=1 |
|
elif choice=="schnell": |
|
speed=0.6 |
|
else: |
|
speed=0.2 |
|
return speed |
|
|
|
def change_stop(choice): |
|
|
|
if choice=="ganz kuez": |
|
stop=0.2 |
|
elif choice=="kuerz": |
|
stop=0.6 |
|
elif choice=="mëttel": |
|
stop=1 |
|
elif choice=="laang": |
|
stop=2.5 |
|
else: |
|
stop=4 |
|
return stop |
|
|
|
def init(): |
|
key_list=['lb_LU-marylux-medium','lb_LU-femaleLOD-medium','lb_LU-androgynous-medium'] |
|
return(gr.update(label="Voice",choices=key_list,value="lb_LU-femaleLOD-medium",interactive=True)) |
|
|
|
def new_load_mod(instr="en_US-joe-medium"): |
|
model=instr |
|
print(f"model: {model}") |
|
lang=instr.split("_")[0] |
|
|
|
print(f"lang: {lang}") |
|
dia=instr.split("-")[0] |
|
|
|
print(f"dia: {dia}") |
|
name=instr.split("-")[1] |
|
|
|
print(f"name: {name}") |
|
style=instr.split("-")[2] |
|
|
|
print(f"style: {style}") |
|
file=f'{instr}.onnx' |
|
print(f"file: {file}") |
|
print(f"Loading model: {file}") |
|
print(f"os.path: {os.path}") |
|
if not os.path.isfile(f'{os.getcwd()}/voices/{file}'): |
|
print(f"Model not found locally") |
|
|
|
m_path= f"https://huggingface.co/mbarnig/lb_rhasspy_piper_tts/resolve/main/{lang}/{dia}/{name}/{style}/{file}" |
|
print(f"m_path: {m_path}") |
|
print("Downloading json...") |
|
json_file=requests.get(f"{m_path}.json") |
|
print("Downloading model...") |
|
mod_file=requests.get(m_path) |
|
with open(f'{os.getcwd()}/voices/{file}','wb') as m: |
|
m.write(mod_file.content) |
|
m.close() |
|
with open(f'{os.getcwd()}/voices/{file}.json','wb') as j: |
|
j.write(json_file.content) |
|
j.close() |
|
pp.json_ob=f'{os.getcwd()}/voices/{file}.json' |
|
print("Model Loaded") |
|
|
|
def load_mod(instr="en_US-joe-medium"): |
|
load_mes=gr.Info(f"""Loading Model...<br>{instr}""",duration=2) |
|
|
|
new_load_mod(instr=instr) |
|
with open(pp.json_ob,'r') as f: |
|
|
|
json_ob=f.read() |
|
load_mes=gr.Info(f"Model Loaded<br>{instr}",duration=2) |
|
return json_ob |
|
|
|
|
|
def save_set(model,length,noise,width,sen_pause): |
|
if not os.path.isdir(f'{os.getcwd()}/saved'): |
|
os.mkdir(f'{os.getcwd()}/saved') |
|
set_json={"model":model,"length":length,"noise":noise,"width":width,"pause":sen_pause} |
|
file_name=f'{model}__{length}__{noise}__{width}__{sen_pause}'.replace(".","_") |
|
with open(f'{os.getcwd()}/saved/{file_name}.json','w') as file: |
|
file.write(json.dumps(set_json,indent=4)) |
|
file.close() |
|
return(f'{os.getcwd()}/saved/{file_name}.json') |
|
|
|
def load_set(set_file): |
|
with open(set_file,'r') as file: |
|
set_json=json.loads(file.read()) |
|
file.close() |
|
return(gr.update(value=set_json['model']),gr.update(value=set_json['length']), |
|
gr.update(value=set_json['noise']),gr.update(value=set_json['width']), |
|
gr.update(value=set_json['pause'])) |
|
|
|
|
|
|
|
def button_on(stream): |
|
if stream==True: |
|
return gr.update(interactive=True,visible=True),gr.update(interactive=False,visible=False) |
|
if stream==False: |
|
return gr.update(interactive=False,visible=False),gr.update(interactive=True,visible=True) |
|
|
|
def clear_aud(): |
|
return None |
|
|
|
with gr.Blocks() as b: |
|
gr.HTML("<h1>Rhasspy Piper LU TTS Streaming</h1>") |
|
with gr.Row(): |
|
with gr.Column(scale=2): |
|
in_txt=gr.Textbox(label="Text",lines=10, value=my_examples) |
|
names=gr.Dropdown() |
|
with gr.Row(): |
|
stream_btn=gr.Button("Stream",interactive=True,visible=True) |
|
sub_btn=gr.Button(interactive=False,visible=False) |
|
cancel_btn=gr.Button("Stop") |
|
out_aud=gr.Audio(streaming=True, autoplay=True) |
|
with gr.Column(scale=1): |
|
with gr.Row(): |
|
with gr.Accordion("Informatiounen", open=False): |
|
gr.Markdown(""" |
|
D'Rhasspy **Piper TTS** Technologie gouf vum Michael Hansen, alias [Synesthesiam](https://github.com/synesthesiam), entwéckelt. |
|
Hien huet och d'Marylux Stëmm trainéiert. Déi zwou aner Lëtzebuergesch Stëmme goufe vum Marco Barnig realiséiert. |
|
|
|
D'Piper Technik berout op enger Ëmwandlung vun Text an **eSpeak-Phonemen**, déi zesumme mat den entspriechenden Audio-Dateien an |
|
engem **neuronale KI-Netz** trainéiert ginn. Déi generéiert TTS-Modeller sinn optiméiert fir Streaming a kënnen a Screenliesmaschinnen, |
|
wéi [NVDA](https://www.nvaccess.org/about-nvda/), agesat ginn. Sie kënnen awer och am Ganze synthetiséiert ginn, |
|
woubäi d'Qualitéit da besser gëtt. |
|
|
|
Verschidde Parameter wéi Liestempo, Sazpaus a Geräischintensitéiten kënnen mat Radioknäpp oder mat Schieberen (am Control-Tab) |
|
agestallt ginn. D'Parameter kënnen och an enger Datei ofgespäichert a reimportéiert ginn.""") |
|
with gr.Row(): |
|
vitess=gr.Radio(label="Liestempo", choices = speeds, value = "normal") |
|
pause=gr.Radio(label="Sazpaus", choices = stops, value = "mëttel") |
|
|
|
with gr.Accordion("Control", open=False): |
|
stream=gr.Checkbox(label="Stream",info="Streaming is fast, but lower quality",value=True,interactive=True) |
|
length=gr.Slider(label="Length", minimum=0.01, maximum=10.0, value=1) |
|
noise=gr.Slider(label="Noise", minimum=0.01, maximum=3.0, value=0.5, visible=True) |
|
width=gr.Slider(label="Noise Width", minimum=0.01, maximum=3.0, value=0.5, visible=True) |
|
sen_pause=gr.Slider(label="Sentence Pause", minimum=0.1, maximum=10.0, value=1, visible=True) |
|
with gr.Tab("Save Settings"): |
|
save_btn=gr.Button("Save") |
|
save_file=gr.File() |
|
with gr.Tab("Load Settings"): |
|
load_file=gr.File() |
|
with gr.Accordion("Model Config", open=False): |
|
json_ob=gr.JSON(label="JSON") |
|
f1=stream.change(button_on,stream,[stream_btn,sub_btn]) |
|
f2=save_btn.click(save_set,[names,length,noise,width,sen_pause],save_file) |
|
f3=load_file.change(load_set,load_file,[names,length,noise,width,sen_pause]) |
|
f4=names.change(load_mod,names,json_ob).then(clear_aud,None,out_aud) |
|
f5=stream_btn.click(clear_aud,None,out_aud) |
|
f6=stream_btn.click(pp.stream_tts,[in_txt,names,length,noise,width,sen_pause],out_aud) |
|
f7=sub_btn.click(clear_aud,None,out_aud) |
|
f8=sub_btn.click(pp.tts,[in_txt,names,length,noise,width,sen_pause],out_aud) |
|
cancel_btn.click(None,None,None,cancels=[f1,f2,f3,f4,f5,f6,f7,f8]) |
|
vitess.change(change_speed, inputs=vitess, outputs=length) |
|
pause.change(change_stop, inputs=pause, outputs=sen_pause) |
|
|
|
b.load(init,None,names) |
|
|
|
|
|
b.launch() |
