Spaces:
Sleeping
Sleeping
| import spaces | |
| import gradio as gr | |
| import edge_tts | |
| import asyncio | |
| import tempfile | |
| import os | |
| # Get all available voices | |
| async def get_voices(): | |
| voices = await edge_tts.list_voices() | |
| return {f"{v['ShortName']} - {v['Locale']} ({v['Gender']})": v['ShortName'] for v in voices} | |
| # Text-to-speech function for a single paragraph | |
| async def paragraph_to_speech(text, voice, rate, pitch): | |
| voice1 = "en-HK-SamNeural - en-HK (Male)" | |
| voice2 = "en-HK-YanNeural - en-HK (Female)" | |
| if not text.strip(): | |
| return None | |
| if text.startswith("FF"): | |
| text2 = text[2:] # Remove the first two characters ("FF") | |
| voice_short_name =voice2.split(" - ")[0] | |
| else: | |
| voice_short_name = voice.split(" - ")[0] | |
| text2=text | |
| rate_str = f"{rate:+d}%" | |
| pitch_str = f"{pitch:+d}Hz" | |
| communicate = edge_tts.Communicate(text2, voice_short_name, rate=rate_str, pitch=pitch_str) | |
| with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp_file: | |
| tmp_path = tmp_file.name | |
| await communicate.save(tmp_path) | |
| return tmp_path | |
| # Main text-to-speech function that processes paragraphs | |
| async def text_to_speech(text, voice, rate, pitch): | |
| if not text.strip(): | |
| return None, gr.Warning("Please enter text to convert.") | |
| if not voice: | |
| return None, gr.Warning("Please select a voice.") | |
| paragraphs = [p for p in text.split("\n\n") if p.strip()] | |
| audio_files = [] | |
| for paragraph in paragraphs: | |
| audio_path = await paragraph_to_speech(paragraph, voice, rate, pitch) | |
| if audio_path: | |
| audio_files.append(audio_path) | |
| if not audio_files: | |
| return None, None # No audio generated | |
| # Combine audio files if there are multiple paragraphs | |
| if len(audio_files) == 1: | |
| return audio_files[0], None | |
| else: | |
| # Simple concatenation for now - consider using a proper audio editing library for smoother transitions | |
| combined_audio_path = tempfile.mktemp(suffix=".mp3") | |
| with open(combined_audio_path, 'wb') as outfile: | |
| for filename in audio_files: | |
| with open(filename, 'rb') as infile: | |
| outfile.write(infile.read()) | |
| os.remove(filename) # Clean up individual files | |
| return combined_audio_path, None | |
| # Gradio interface function | |
| def tts_interface(text, voice, rate, pitch): | |
| audio, warning = asyncio.run(text_to_speech(text, voice, rate, pitch)) | |
| return audio, warning | |
| # Create Gradio application | |
| import gradio as gr | |
| async def create_demo(): | |
| voices = await get_voices() | |
| description = """ | |
| Experience the power of Voicecloning.be for text-to-speech conversion. | |
| Enter your text, select a voice, and adjust the speech rate and pitch. | |
| The application will process your text paragraph by paragraph (separated by two blank lines). | |
| """ | |
| demo = gr.Interface( | |
| fn=tts_interface, | |
| inputs=[ | |
| gr.Textbox(label="Input Text", lines=5, placeholder="Separate paragraphs with two blank lines."), | |
| gr.Dropdown(choices=[""] + list(voices.keys()), label="Select Voice", value=""), | |
| gr.Slider(minimum=-50, maximum=50, value=0, label="Speech Rate Adjustment (%)", step=1), | |
| gr.Slider(minimum=-20, maximum=20, value=0, label="Pitch Adjustment (Hz)", step=1) | |
| ], | |
| outputs=[ | |
| gr.Audio(label="Generated Audio", type="filepath"), | |
| gr.Markdown(label="Warning", visible=False) | |
| ], | |
| title="Voicecloning.be Text-to-Speech (Paragraph by Paragraph)", | |
| description=description, | |
| article="Process text paragraph by paragraph for smoother output.", | |
| analytics_enabled=False, | |
| allow_flagging=False | |
| ) | |
| return demo | |
| # Run the application | |
| if __name__ == "__main__": | |
| demo = asyncio.run(create_demo()) | |
| demo.launch() |