File size: 8,462 Bytes
bb63470 00ada4a db42dba 8b6ea6c e006b08 8b6ea6c bb63470 db42dba 6045e4a e006b08 6045e4a 00ada4a 7ec9920 9d47e7f 6045e4a 8b6ea6c 7ec9920 8b6ea6c db42dba 9d47e7f 6045e4a 9d47e7f 6045e4a 9d47e7f e006b08 8b6ea6c 9d47e7f e006b08 9d47e7f e006b08 9d47e7f db42dba 9d47e7f 6045e4a 9d47e7f db42dba 6045e4a 9d47e7f e006b08 9d47e7f 6045e4a e006b08 db42dba 8b6ea6c 9d47e7f e006b08 6045e4a e006b08 6045e4a e006b08 6045e4a e006b08 6045e4a e006b08 8b6ea6c e006b08 db42dba 6045e4a e006b08 6045e4a 8b6ea6c db42dba 7ec9920 e006b08 6045e4a 8b6ea6c db42dba 7ec9920 9d47e7f e006b08 9d47e7f db42dba 6045e4a 9d47e7f 8b6ea6c db42dba 7ec9920 6045e4a e006b08 db42dba e006b08 7ec9920 9f7512d bb63470 9d47e7f |
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 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 |
import gradio as gr
import numpy as np
import librosa
from transformers import pipeline
from datetime import datetime
import os
import requests
# Inference API 설정
API_URL = "https://api-inference.huggingface.co/models/stabilityai/stable-diffusion-xl-base-1.0"
headers = {"Authorization": "Bearer hf_..."} # 여기에 HuggingFace 토큰 입력
# AI 모델 초기화
speech_recognizer = pipeline(
"automatic-speech-recognition",
model="kresnik/wav2vec2-large-xlsr-korean"
)
emotion_classifier = pipeline(
"audio-classification",
model="MIT/ast-finetuned-speech-commands-v2"
)
text_analyzer = pipeline(
"sentiment-analysis",
model="nlptown/bert-base-multilingual-uncased-sentiment"
)
def create_interface():
with gr.Blocks(theme=gr.themes.Soft()) as app:
state = gr.State({
"user_name": "",
"reflections": [],
"voice_analysis": None,
"final_prompt": ""
})
def generate_image_from_prompt(prompt):
"""HuggingFace Inference API를 통한 이미지 생성"""
try:
response = requests.post(API_URL, headers=headers, json={
"inputs": prompt,
"parameters": {
"negative_prompt": "ugly, blurry, poor quality, distorted",
"num_inference_steps": 30,
"guidance_scale": 7.5
}
})
if response.status_code == 200:
return response.content # 바이너리 이미지 데이터 반환
else:
return None
except Exception as e:
print(f"Error generating image: {e}")
return None
# 헤더
header = gr.Markdown("# 디지털 굿판")
user_display = gr.Markdown("")
with gr.Tabs() as tabs:
# 입장
with gr.Tab("입장"):
gr.Markdown("""# 디지털 굿판에 오신 것을 환영합니다""")
name_input = gr.Textbox(label="이름을 알려주세요")
start_btn = gr.Button("여정 시작하기")
# 청신
with gr.Tab("청신"):
with gr.Row():
# 절대 경로로 변경
audio_path = os.path.abspath(os.path.join("assets", "main_music.mp3"))
audio = gr.Audio(
value=audio_path,
type="filepath",
label="온천천의 소리",
interactive=False,
autoplay=True
)
with gr.Column():
reflection_input = gr.Textbox(
label="현재 순간의 감상을 적어주세요",
lines=3
)
save_btn = gr.Button("감상 저장하기")
reflections_display = gr.Dataframe(
headers=["시간", "감상", "감정 분석"],
label="기록된 감상들"
)
# 기원
with gr.Tab("기원"):
gr.Markdown("## 기원 - 목소리로 전하기")
with gr.Row():
with gr.Column():
voice_input = gr.Audio(
label="나누고 싶은 이야기를 들려주세요",
sources=["microphone"],
type="filepath",
interactive=True
)
clear_btn = gr.Button("녹음 지우기")
with gr.Column():
transcribed_text = gr.Textbox(
label="인식된 텍스트",
interactive=False
)
voice_emotion = gr.Textbox(
label="음성 감정 분석",
interactive=False
)
text_emotion = gr.Textbox(
label="텍스트 감정 분석",
interactive=False
)
analyze_btn = gr.Button("분석하기")
# 이벤트 핸들러 추가
generate_btn.click(
fn=generate_image_from_prompt,
inputs=[final_prompt],
outputs=[result_image]
)
# 송신 탭 부분 수정
with gr.Tab("송신"):
gr.Markdown("## 송신 - 시각화 결과")
with gr.Column():
final_prompt = gr.Textbox(
label="생성된 프롬프트",
interactive=False,
lines=3
)
generate_btn = gr.Button("이미지 생성하기")
result_image = gr.Image(
label="생성된 이미지",
type="pil"
)
def clear_voice_input():
"""음성 입력 초기화"""
return None
def analyze_voice(audio_path, state):
"""음성 분석"""
if audio_path is None:
return state, "음성을 먼저 녹음해주세요.", "", "", ""
try:
# 오디오 로드
y, sr = librosa.load(audio_path, sr=16000)
# 음성 인식
transcription = speech_recognizer(y)
text = transcription["text"]
# 감정 분석
voice_emotions = emotion_classifier(y)
text_sentiment = text_analyzer(text)[0]
# 프롬프트 생성
prompt = generate_prompt(text, voice_emotions[0], text_sentiment)
return (
state,
text,
f"음성 감정: {voice_emotions[0]['label']} ({voice_emotions[0]['score']:.2f})",
f"텍스트 감정: {text_sentiment['label']} ({text_sentiment['score']:.2f})",
prompt
)
except Exception as e:
return state, f"오류 발생: {str(e)}", "", "", ""
def generate_prompt(text, voice_emotion, text_sentiment):
"""프롬프트 생성"""
emotion_colors = {
"happy": "따뜻한 노란색과 주황색",
"sad": "깊은 파랑색과 보라색",
"angry": "강렬한 빨강색과 검정색",
"neutral": "부드러운 회색과 베이지색"
}
color = emotion_colors.get(voice_emotion['label'], "자연스러운 색상")
prompt = f"한국 전통 민화 스타일의 추상화, {color} 사용. "
prompt += f"음성의 감정({voice_emotion['label']})과 텍스트의 감정({text_sentiment['label']})이 조화를 이루며, "
prompt += f"음성의 특징을 반영한 동적인 구도. 발화 내용: '{text}'"
return prompt
def save_reflection(text, state):
"""감상 저장"""
if not text.strip():
return state, state["reflections"]
current_time = datetime.now().strftime("%H:%M:%S")
sentiment = text_analyzer(text)[0]
new_reflection = [current_time, text, f"{sentiment['label']} ({sentiment['score']:.2f})"]
if "reflections" not in state:
state["reflections"] = []
state["reflections"].append(new_reflection)
return state, state["reflections"]
# 이벤트 연결
start_btn.click(
fn=lambda name: (f"# 환영합니다, {name}님의 디지털 굿판", gr.update(selected="청신")),
inputs=[name_input],
outputs=[user_display, tabs]
)
save_btn.click(
fn=save_reflection,
inputs=[reflection_input, state],
outputs=[state, reflections_display]
)
clear_btn.click(
fn=clear_voice_input,
inputs=[],
outputs=[voice_input]
)
analyze_btn.click(
fn=analyze_voice,
inputs=[voice_input, state],
outputs=[state, transcribed_text, voice_emotion, text_emotion, final_prompt]
)
return app
if __name__ == "__main__":
demo = create_interface()
demo.launch() |