Spaces:
Sleeping
Sleeping
| import io | |
| import json | |
| import os | |
| import requests | |
| import urllib | |
| import gradio as gr | |
| import numpy as np | |
| from scipy.io import wavfile | |
| base_url = "https://api.sandbox.deepgram.com/tts" | |
| token_str = os.environ['DG_TOKEN'] | |
| def tts_fn(text, prompt_audio, prompt_seconds, inference_steps, inference_temperature, pitch_steps): | |
| texts = [text] | |
| sr = prompt_audio[0] | |
| prompt_audio = prompt_audio[1].astype(np.float32, order='C') / 32768.0 | |
| byte_io = io.BytesIO(bytes()) | |
| wavfile.write(byte_io, sr, prompt_audio) | |
| prompt_audio_bytes = byte_io.read() | |
| params={'synthesize': 'true', 'text': urllib.parse.quote(text), 'pitch_steps': int(pitch_steps), 'soundstorm_steps': inference_steps, 'temperature': inference_temperature, 'prompt_seconds': prompt_seconds} | |
| response = requests.post(base_url, data=prompt_audio_bytes, params=params, headers={'Authorization': f'Token {token_str}'}) | |
| try: | |
| with open('result.wav', 'wb') as f: | |
| for chunk in response.iter_content(chunk_size=1024): | |
| if chunk: f.write(chunk) | |
| sample_rate, audio = wavfile.read('result.wav') | |
| print(audio.dtype) | |
| audio = (audio / 1.414 * 32767).astype(np.int16) | |
| print(audio.dtype) | |
| except Exception: | |
| print(response) | |
| return (sample_rate, audio) | |
| demo_files = ['demo_files/man.wav', 'demo_files/woman.wav', 'demo_files/man_2.wav', 'demo_files/woman_2.wav', 'demo_files/man_3.wav', 'demo_files/woman_3.wav', 'demo_files/woman_4.wav', 'demo_files/meditation.wav'] | |
| app = gr.Blocks() | |
| with app: | |
| with gr.Tab("TTS MVP"): | |
| with gr.Row(): | |
| with gr.Column(): | |
| pangram = "The beige hue on the waters of the loch impressed all, including the French queen, before she heard that symphony again, just as young Arthur wanted." | |
| cherry = "Your request has been processed and the audio is ready for playback." | |
| textbox = gr.TextArea(label="Text", placeholder="Type a sentence here", value=cherry) | |
| prompt_audio = gr.Audio(label="Prompt Audio", source='upload') | |
| examples = gr.Examples(label='Sample Speakers', examples=demo_files, inputs=prompt_audio) | |
| # speed = gr.Slider(minimum=0.0, maximum=2.0, value=1.1, step=0.1, label="Speed") | |
| # variability = gr.Slider(minimum=0.0, maximum=1.0, value=0.7, step=0.1, label="Variability") | |
| inference_steps = gr.Slider(minimum=1, maximum=32, value=1, step=1, label="Inference Steps: quality vs latency tradeoff. Results are sometimes unstable for values >1.") | |
| inference_temperature = gr.Slider(minimum=0.0, maximum=1.0, value=0.9, step=0.05, label="Temperature: fidelity vs variability tradeoff") | |
| prompt_seconds = gr.Slider(minimum=1.0, maximum=10.0, value=3.0, step=1.0, label="Use first N seconds of prompt audio") | |
| pitch_steps = gr.Slider(minimum=-24, maximum=24, value=0, step=1, label="Pitch Steps: 12 to an octave") | |
| with gr.Column(): | |
| audio_output = gr.Audio(label="Output Audio", elem_id='tts-audio') | |
| btn = gr.Button("Generate") | |
| btn.click(tts_fn, inputs=[textbox, prompt_audio, prompt_seconds, inference_steps, inference_temperature, pitch_steps], outputs=[audio_output]) | |
| app.launch() |