app.py

import sys
import os
from fastapi import Request
# By using XTTS you agree to CPML license https://coqui.ai/cpml
os.environ["COQUI_TOS_AGREED"] = "1"

import gradio as gr
from TTS.api import TTS
from TTS.utils.manage import ModelManager
model_names = TTS().list_models()
print(model_names.__dict__)
print(model_names.__dir__())
model_name = "tts_models/multilingual/multi-dataset/xtts_v2" # move in v2, since xtts_v1 is generated keyerror, I guess you can select it with old github's release.

#m = ModelManager().download_model(model_name)
#print(m)
m = model_name

tts = TTS(model_name, gpu=False)
tts.to("cpu") # no GPU or Amd
#tts.to("cuda") # cuda only


def predict(prompt, language, audio_file_pth, mic_file_path, use_mic, agree, request: gr.Request):
    """
    Due to the large number of abuses observed in the console logs, I have been forced to integrate the ‘display of additional information’ relating to the use of this space.
    I have been obliged to include a ‘display of additional information’ relating to the use of this area.
    As a reminder, sending illegal content (se*ual, offensive or threatening content) in any language is of course FORBIDDEN.
    language, is of course FORBIDDEN. I cannot be held responsible for those who violate a strictly [ETHICAL and MORAL] use of this space.
    use of this model.
    """

    try:
        if request:
            print("Request obj:", request.__dict__)
            print("= = = = = = = = = = = =")
            print("Request headers dictionary:", request.headers)
            print("IP address:", request.client.host)
            try:
                print("Cookies:", request.__dict__.get("cookie"))
            except:
                print("no Cookies")
            print("Query parameters:", dict(request.query_params))
            print()
        else:
            print("no such Request..")
    except Exception as e:
        print(e)

    if agree == True:
        if use_mic == True:
            if mic_file_path is not None:
                speaker_wav=mic_file_path
            else:
                gr.Warning("Please record your voice with Microphone, or uncheck Use Microphone to use reference audios")
                return (
                    None,
                    #None,
                ) 
                
        else:
            speaker_wav=audio_file_pth

        if len(prompt)<2:
            gr.Warning("Please give a longer prompt text")
            return (
                   # None,
                    None,
                )
        if len(prompt)>10000:
            gr.Warning("Text length limited to 10000 characters for this demo, please try shorter text")
            return (
                  #  None,
                    None,
                )  
        try:
            if language == "fr":
                if m.find("your") != -1:
                    language = "fr-fr"
            if m.find("/fr/") != -1:
                language = None
            tts.tts_to_file(
                text=prompt,
                file_path="output.wav",
                speaker_wav=speaker_wav,
                language=language
            )
        except RuntimeError as e :
            if "device-assert" in str(e):
                # cannot do anything on cuda device side error, need tor estart
                gr.Warning("Unhandled Exception encounter, please retry in a minute")
                print("Cuda device-assert Runtime encountered need restart")
                sys.exit("Exit due to cuda device-assert")
            else:
                raise e
            
        return (
            #gr.make_waveform(
                #audio="output.wav",
            #),
            "output.wav",
        )
    else:
        gr.Warning("Please accept the Terms & Condition!")
        return (
               # None,
                None,
            ) 


title = "XTTS Glz's remake (Fonctional Text-2-Speech)"

description = f"""
<a href="https://huggingface.co/coqui/XTTS-v1">XTTS</a> is a Voice generation model that lets you clone voices into different languages by using just a quick 3-second audio clip. 
<br/>
XTTS is built on previous research, like Tortoise, with additional architectural innovations and training to make cross-language voice cloning and multilingual speech generation possible. 
<br/>
This is the same model that powers our creator application <a href="https://coqui.ai">Coqui Studio</a> as well as the <a href="https://docs.coqui.ai">Coqui API</a>. In production we apply modifications to make low-latency streaming possible.
<br/>
Leave a star on the Github <a href="https://github.com/coqui-ai/TTS">TTS</a>, where our open-source inference and training code lives.
<br/>
<p>For faster inference without waiting in the queue, you should duplicate this space and upgrade to GPU via the settings.
<br/>
<a href="https://huggingface.co/spaces/coqui/xtts?duplicate=true">
<img style="margin-top: 0em; margin-bottom: 0em" src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>
</p>
<p>
<br>
Due to the large number of abuses observed in the console's logs, ‘additional information’ on the use of this space has been added.
<br>
As a reminder, sending illegal content (se*ual, offensive or threatening content) in any language is, of course, FORBIDDEN.
<br>
The authors of this space cannot be held responsible for those who violate the strictly [ETHICAL AND MORAL] use of this model.
<br>
</p>
"""

article = """
<div style='margin:20px auto;'>
<p>By using this demo you agree to the terms of the Coqui Public Model License at https://coqui.ai/cpml</p>
</div>
"""
examples = [
    [
        "Hello, World !, here is an example of light voice cloning. Try to upload your best audio samples quality",
        ["English", "en"],
        "examples/female.wav",
        None,
        False,
        True,
    ],
    [
        "Je suis un lycéen français de 17 ans, passioner par la Cyber-Sécuritée et les models d'IA.",
        ["French", "fr"],
        "examples/female.wav",
        None,
        False,
        True,
    ],
    [
        "Als ich sechs war, sah ich einmal ein wunderbares Bild",
        ["German", "de"],
        "examples/female.wav",
        None,
        False,
        True,
    ],
    [
        "Cuando tenía seis años, vi una vez una imagen magnífica",
        ["Spanish", "es"],
        "examples/female.wav",
        None,
        False,
        True,
    ],
    [
        "Quando eu tinha seis anos eu vi, uma vez, uma imagem magnífica",
        ["Portuguese", "pt"],
        "examples/female.wav",
        None,
        False,
        True,
    ],
    [
        "Kiedy miałem sześć lat, zobaczyłem pewnego razu wspaniały obrazek",
        ["Polish", "pl"],
        "examples/female.wav",
        None,
        False,
        True,
    ],
    [
        "Un tempo lontano, quando avevo sei anni, vidi un magnifico disegno",
        ["Italian", "it"],
        "examples/female.wav",
        None,
        False,
        True,
    ],
    [
        "Bir zamanlar, altı yaşındayken, muhteşem bir resim gördüm",
        ["Turkish", "tr"],
        "examples/female.wav",
        None,
        False,
        True,
    ],
    [
        "Когда мне было шесть лет, я увидел однажды удивительную картинку",
        ["Russian", "ru"],
        "examples/female.wav",
        None,
        False,
        True,
    ],
    [
        "Toen ik een jaar of zes was, zag ik op een keer een prachtige plaat",
        ["Dutch", "nl"],
        "examples/female.wav",
        None,
        False,
        True,
    ],
    [
        "Když mi bylo šest let, viděl jsem jednou nádherný obrázek",
        ["Czech", "cs"],
        "examples/female.wav",
        None,
        False,
        True,
    ],
    [
        "当我还只有六岁的时候， 看到了一副精彩的插画",
        ["Mandarin Chinese", "zh-cn"],
        "examples/female.wav",
        None,
        False,
        True,
    ],
]



gr.Interface(
    fn=predict,
    inputs=[
        gr.Textbox(
            label="Text Prompt",
            info="One or two sentences at a time is better",
            value="Hello, World !, here is an example of light voice cloning. Try to upload your best audio samples quality",
        ),
        gr.Dropdown(
            label="Language",
            info="Select an output language for the synthesised speech",
            choices=[
                    ["Arabic", "ar"],
                    ["Portuguese", "pt"],
                    ["Mandarin Chinese", "zh-cn"],
                    ["Czech", "cs"],
                    ["Dutch", "nl"],
                    ["English", "en"],
                    ["French", "fr"],
                    ["German", "de"],
                    ["Italian", "it"],
                    ["Polish", "pl"],
                    ["Russian", "ru"],
                    ["Spanish", "es"],
                    ["Turkish", "tr"],
            ],
            value="en",
        ),
        gr.Audio(
            label="Reference Audio",
            #info="Click on the ✎ button to upload your own target speaker audio",
            type="filepath",
            value="examples/female.wav",
        ),
        gr.Audio(sources=["microphone"],
                 type="filepath",
                 #info="Use your microphone to record audio",
                 label="Use Microphone for Reference"),
        gr.Checkbox(label="Check to use Microphone as Reference",
                    value=False,
                    info="Notice: Microphone input may not work properly under traffic",),
        gr.Checkbox(
            label="Agree",
            value=True,
            info="I agree to the terms of the Coqui Public Model License at https://coqui.ai/cpml",
        ),
    ],
    outputs=[
        #gr.Video(label="Waveform Visual"),
        gr.Audio(label="Synthesised Audio"),
    ],
    title=title,
    description=description,
    article=article,
    #examples=examples, # cause errors about False is not an audio
).queue().launch(debug=True)