| import gradio as gr | |
| from TTS.api import TTS | |
| tts = TTS("tts_models/multilingual/multi-dataset/xtts_v1", gpu=True) | |
| def predict(prompt, language, audio_file_pth): | |
| tts.tts_to_file(text=prompt, | |
| file_path="output.wav", | |
| speaker_wav=audio_file_pth, | |
| language=language) | |
| return gr.make_waveform(audio="output.wav",) | |
| title = "XTTS: MVP" | |
| gr.Interface( | |
| fn=predict, | |
| inputs=[ | |
| gr.Textbox(label="Prompt", info = "One or two sentences at a time is better* (max: 10)", placeholder = "It took me quite a long time to develop a voice, and now that I have it I'm not going to be silent.",), | |
| gr.Dropdown(choices=["en", "es", "fr", "de", "it", "pt", "pl", "tr", "ru", "nl", "cz", "ar", "zh"], max_choices=1), | |
| gr.Audio(label="Upload Speaker WAV", type="filepath"), | |
| ], | |
| outputs=[ | |
| gr.Video(label="Synthesised Speech"), | |
| ], | |
| title=title, | |
| ).launch(debug=True) |