Delete app.bak.py

app.bak.py

@@ -1,160 +0,0 @@
-import gradio as gr
-import numpy as np
-import torch
-import torch.nn.functional as F
-from pathlib import Path
-
-from TTS.api import TTS
-from TTS.utils.manage import ModelManager
-
-
-title = ""
-description = """"""
-article = """"""
-
-device = "cuda" if torch.cuda.is_available() else "cpu"
-GPU = device == "cuda"
-INT16MAX = np.iinfo(np.int16).max
-
-model_ids = ModelManager(verbose=False).list_models()
-model_tts_ids = [model for model in model_ids if 'tts_models' in model and ('/multilingual/' in model or '/en/' in model)]
-model_voc_ids = [model for model in model_ids if 'vocoder_models' in model and ('/universal/' in model or '/en/' in model)]
-model_vc_ids = [model for model in model_ids if 'voice_conversion_models' in model and ('/multilingual/' in model or '/en/' in model)]
-examples_pt = 'examples'
-allowed_extentions = ['.mp3', '.wav']
-examples = {f.name: f for f in Path(examples_pt).glob('*') if f.suffix in allowed_extentions}
-verse = """Mary had a little lamb,
-Its fleece was white as snow.
-Everywhere the child went,
-The little lamb was sure to go."""
-
-
-
-def on_model_tts_select(model_name, tts_var):
-    if tts_var is None or tts_var.model_name != model_name:
-        print(f'Loading TTS model from {model_name}')
-        tts_var = TTS(model_name=model_name, progress_bar=False, gpu=GPU)
-    else:
-        print(f'Passing through TTS model {tts_var.model_name}')
-    languages = tts_var.languages if tts_var.is_multi_lingual else ['']
-    speakers = [s.replace('\n', '-n') for s in tts_var.speakers] if tts_var.is_multi_speaker else [''] # there's weird speaker formatting
-    language = languages[0]
-    speaker = speakers[0]
-    return tts_var, gr.update(choices=languages, value=language, interactive=tts_var.is_multi_lingual),\
-                gr.update(choices=speakers, value=speaker, interactive=tts_var.is_multi_speaker)
-
-
-def on_model_vc_select(model_name, vc_var):
-    if vc_var is None or vc_var.model_name != model_name:
-        print(f'Loading voice conversion model from {model_name}')
-        vc_var = TTS(model_name=model_name, progress_bar=False, gpu=GPU)
-    else:
-        print(f'Passing through voice conversion model {vc_var.model_name}')
-    return vc_var
-
-
-def on_voicedropdown(x):
-    return examples[x]
-
-
-def text_to_speech(text, tts_model, language, speaker, target_wav, use_original_voice):
-    if len(text.strip()) == 0 or tts_model is None or (target_wav is None and not use_original_voice):
-        return (16000, np.zeros(0).astype(np.int16))
-    
-    sample_rate = tts_model.synthesizer.output_sample_rate
-    if tts_model.is_multi_speaker:
-        speaker = {s.replace('\n', '-n'): s for s in tts_model.speakers}[speaker] # there's weird speaker formatting
-    print(f'model: {tts_model.model_name}\nlanguage: {language}\nspeaker: {speaker}')
-    
-    language = None if language == '' else language
-    speaker = None if speaker == '' else speaker
-    if use_original_voice:
-        print('Using original voice')
-        speech = tts_model.tts(text, language=language, speaker=speaker)       
-    elif tts_model.synthesizer.tts_model.speaker_manager:
-        print('voice cloning with the tts')
-        speech = tts_model.tts(text, language=language, speaker_wav=target_wav)
-    else:
-        print('voice cloning with the voice conversion model')
-        speech = tts_model.tts_with_vc(text, language=language, speaker_wav=target_wav)
-
-    speech = (np.array(speech) * INT16MAX).astype(np.int16)
-    return (sample_rate, speech)
-
-
-def voice_clone(vc_model, source_wav, target_wav):
-    print(f'model: {vc_model.model_name}\nsource_wav: {source_wav}\ntarget_wav: {target_wav}')
-    sample_rate = vc_model.voice_converter.output_sample_rate
-    if vc_model is None or source_wav is None or target_wav is None:
-        return (sample_rate, np.zeros(0).astype(np.int16))
-
-    speech = vc_model.voice_conversion(source_wav=source_wav, target_wav=target_wav)
-    speech = (np.array(speech) * INT16MAX).astype(np.int16)
-    return (sample_rate, speech)
-
-
-with gr.Blocks() as demo:
-    tts_model = gr.State(None)
-    vc_model = gr.State(None)
-    def activate(*args):
-        return gr.update(interactive=True) if len(args) == 1 else [gr.update(interactive=True)] * len(args)
-    def deactivate(*args):
-        return gr.update(interactive=False) if len(args) == 1 else [gr.update(interactive=False)] * len(args)
-
-    gr.Markdown(description)
-
-    with gr.Row(equal_height=True):
-        with gr.Column(scale=5, min_width=50):
-            model_tts_dropdown = gr.Dropdown(model_tts_ids, value=model_tts_ids[3], label='Text-to-speech model', interactive=True)
-        with gr.Column(scale=1, min_width=10):
-                language_dropdown = gr.Dropdown(None, value=None, label='Language', interactive=False, visible=True)
-        with gr.Column(scale=1, min_width=10):
-                speaker_dropdown = gr.Dropdown(None, value=None, label='Speaker', interactive=False, visible=True)
-        with gr.Column(scale=5, min_width=50):
-            with gr.Row(equal_height=True):
-#                 model_vocoder_dropdown = gr.Dropdown(model_voc_ids, label='Select vocoder model', interactive=True)
-                model_vc_dropdown = gr.Dropdown(model_vc_ids, value=model_vc_ids[0], label='Voice conversion model', interactive=True)
-                
-    with gr.Accordion("Target voice", open=False) as accordion:
-        gr.Markdown("Upload target voice...")
-        with gr.Row(equal_height=True):
-            voice_upload = gr.Audio(label='Upload target voice', source='upload', type='filepath')
-            voice_dropdown = gr.Dropdown(examples, label='Examples', interactive=True)
-
-    with gr.Row(equal_height=True):
-        with gr.Column(scale=2):
-            with gr.Row(equal_height=True):
-                with gr.Column():
-                    text_to_convert = gr.Textbox(verse)
-                    orig_voice = gr.Checkbox(label='Use original voice')
-                voice_to_convert = gr.Audio(label="Upload voice to convert", source='upload', type='filepath')
-            with gr.Row(equal_height=True):
-                button_text = gr.Button('Text to speech', interactive=True)
-                button_audio = gr.Button('Convert audio', interactive=True)
-    with gr.Row(equal_height=True):
-        speech = gr.Audio(label='Converted Speech', type='numpy', visible=True, interactive=False) 
-        
-    # actions
-    model_tts_dropdown.change(deactivate, [button_text, button_audio], [button_text, button_audio]).\
-        then(fn=on_model_tts_select, inputs=[model_tts_dropdown, tts_model], outputs=[tts_model, language_dropdown, speaker_dropdown]).\
-        then(activate, [button_text, button_audio], [button_text, button_audio])
-    model_vc_dropdown.change(deactivate, [button_text, button_audio], [button_text, button_audio]).\
-        then(fn=on_model_vc_select, inputs=[model_vc_dropdown, vc_model], outputs=vc_model).\
-        then(activate, [button_text, button_audio], [button_text, button_audio])
-    voice_dropdown.change(deactivate, [button_text, button_audio], [button_text, button_audio]).\
-        then(fn=on_voicedropdown, inputs=voice_dropdown, outputs=voice_upload).\
-        then(activate, [button_text, button_audio], [button_text, button_audio])
-    
-    button_text.click(deactivate, [button_text, button_audio], [button_text, button_audio]).\
-        then(fn=on_model_tts_select, inputs=[model_tts_dropdown, tts_model], outputs=[tts_model, language_dropdown, speaker_dropdown]).\
-        then(fn=text_to_speech, inputs=[text_to_convert, tts_model, language_dropdown, speaker_dropdown, voice_upload, orig_voice], 
-             outputs=speech).\
-        then(activate, [button_text, button_audio], [button_text, button_audio])
-
-    button_audio.click(deactivate, [button_text, button_audio], [button_text, button_audio]).\
-        then(fn=on_model_vc_select, inputs=[model_vc_dropdown, vc_model], outputs=vc_model).\
-        then(fn=voice_clone, inputs=[vc_model, voice_to_convert, voice_upload], outputs=speech).\
-        then(activate, [button_text, button_audio], [button_text, button_audio])
-    
-    gr.HTML(article)
-demo.launch(share=False)