Hyeonseo's picture
Create app.py
e1c7b15 verified
raw
history blame
2.59 kB
import gradio as gr
from openai import OpenAI
import re
from konlpy.tag import Kkma
from TextUtil.digit2text import digit2txt, NNGdigit2txt, CSign2txt
def process_txt(text):
kkma = Kkma()
result = ""
pattern = re.compile(r'([κ°€-힣]+)|([a-zA-Z.]+)|(\d[\d,.]*)|(\$|€|Β£|Β₯|οΏ¦)|(\s+)')
matches = pattern.finditer(text)
for match in matches:
if match.group(1): # Korean part
result += match.group(1)
elif match.group(2):
result += match.group(2)
elif match.group(3): # Number part
end_index = match.end(3)
# NNG Case
next_word = kkma.pos(text[end_index:])[0]
if next_word[1] == "NNG" and next_word[0] not in ['λ‹¬λŸ¬', '유료', 'νŒŒμš΄λ“œ', 'μ—”', '원']:
result += NNGdigit2txt(match.group(3).replace(',', ''))
else:
result += digit2txt(match.group(3).replace(',', ''))
elif match.group(4): # Currency symbol part
result += CSign2txt(match.group(4))
elif match.group(5): # Space part
result += match.group(5)
return result
def generate_audio(api_key, file, model, voice):
# OpenAI ν΄λΌμ΄μ–ΈνŠΈ μ΄ˆκΈ°ν™” (μ‚¬μš©μž μž…λ ₯ API ν‚€ μ‚¬μš©)
client = OpenAI(api_key=api_key)
# 파일 읽기
text = file.decode("utf-8")
# ν…μŠ€νŠΈ 처리
text = process_txt(text)
print(text)
# TTS μš”μ²­
response = client.audio.speech.create(
model=model,
voice=voice,
input=text
)
# MP3 파일둜 μ €μž₯
f_name = "generated_audio"
speech_file_path = f"{f_name}.mp3"
response.stream_to_file(speech_file_path)
return speech_file_path
# Gradio μΈν„°νŽ˜μ΄μŠ€ μ •μ˜
iface = gr.Interface(
fn=generate_audio,
inputs=[
gr.Text(label="Enter OpenAI API Key"),
gr.File(label="Upload Text File", type="binary"),
gr.Radio(choices=["tts-1", "tts-1-hd"], label="Model"),
gr.Radio(choices=["alloy", "echo", "fable", "onyx", "nova", "shimmer"], label="Voice")
],
outputs=gr.File(label="Download MP3 File"),
title="Text-to-Speech Converter (Korean Digit2Text)",
description="Upload a text file and enter your OpenAI API key to convert it into speech using OpenAI's Text-to-Speech models.<br>*ν•΄λ‹Ή μ„œλΉ„μŠ€λŠ” ν•œκ΅­μ–΄μ— λ§žμΆ€ν™”λ˜μ–΄ μžˆμŠ΅λ‹ˆλ‹€. <br>*ν•œκ΅­μ–΄ 숫자 발음 λ³€ν™˜μ„ 톡해 더 μ •ν™•ν•œ 숫자 TTSλ₯Ό κ°€λŠ₯ν•˜κ²Œ ν•©λ‹ˆλ‹€.<br>*μ˜ˆμ‹œ: 50,000$ -> μ˜€λ§Œλ‹¬λŸ¬, 5가지 -> 닀섯가지, 99ꢌ -> μ•„ν”μ•„ν™‰κΆŒ"
)