Spaces:
Sleeping
Sleeping
File size: 1,355 Bytes
4300fed 610f79e 4300fed 61c12f6 7ffbb2d 4300fed 532dc11 610f79e 61c12f6 610f79e 1821dd9 61c12f6 1821dd9 61c12f6 3c963f8 610f79e 61c12f6 3c963f8 61c12f6 610f79e 4300fed 610f79e 4300fed 610f79e 4300fed 610f79e 70cbf96 881961f 532dc11 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 |
import gradio as gr
import spaces
import os, torch, io
import json
os.system('python -m unidic download')
# print("Make sure you've downloaded unidic (python -m unidic download) for this WebUI to work.")
from melo.api import TTS
import tempfile
@spaces.GPU
def synthesize(conversation_text, speed, progress=gr.Progress()):
speed = 1.0
device = 'cuda' if torch.cuda.is_available() else 'cpu'
models = {
'EN': TTS(language='EN', device=device),
}
speakers = ['EN-US', 'EN-Default']
final_audio = io.BytesIO()
conversation = json.loads(conversation_text)
for i, turn in enumerate(conversation["conversation"]):
bio = io.BytesIO()
text = turn["text"]
speaker = speakers[i % 2]
speaker_id = models['EN'].hps.data.spk2id[speaker]
models['EN'].tts_to_file(text, speaker_id, bio, speed=speed, pbar=progress.tqdm, format='wav')
final_audio.write(bio.getvalue())
return final_bio.getvalue()
with gr.Blocks() as demo:
gr.Markdown('# Article to Podcast')
with gr.Group():
text = gr.Textbox(label="Article Link")
btn = gr.Button('Podcasitfy', variant='primary')
aud = gr.Audio(interactive=False)
btn.click(synthesize, inputs=[text], outputs=[aud])
demo.queue(api_open=True, default_concurrency_limit=10).launch(show_api=True)
|