Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,304 +1,352 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
# این اسکریپت دو نقش دارد
|
| 5 |
-
# 1. یک UI Gradio برای تبدیل ویدیو به MP3 و استخراج متن
|
| 6 |
-
# 2. یک API ساده (Flask) برای دریافت وضعیت و نهاییسازی خروجی
|
| 7 |
-
# 3. نگهداری یک دیکشنری `session_map` بهصورت thread‑safe
|
| 8 |
-
# تا `token` تولیدشده در PHP را با `session_hash` Gradio مطابقت دهیم
|
| 9 |
-
# ─────────────────────────────────────────────────────────────────────
|
| 10 |
import os
|
| 11 |
-
import uuid
|
| 12 |
-
import json
|
| 13 |
import shutil
|
| 14 |
-
import tempfile
|
| 15 |
-
import warnings
|
| 16 |
-
import time
|
| 17 |
-
import threading
|
| 18 |
-
import yt_dlp
|
| 19 |
import speech_recognition as sr
|
| 20 |
from pydub import AudioSegment
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
import
|
|
|
|
|
|
|
| 25 |
|
| 26 |
warnings.filterwarnings("ignore")
|
| 27 |
|
| 28 |
-
#
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
#
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
recognizer = sr.Recognizer()
|
| 44 |
recognizer.energy_threshold = 300
|
| 45 |
recognizer.dynamic_energy_threshold = True
|
| 46 |
recognizer.pause_threshold = 0.8
|
| 47 |
-
|
| 48 |
full_text = []
|
| 49 |
bad_chunks = 0
|
| 50 |
total_chunks = 0
|
| 51 |
temp_wav_dir = tempfile.mkdtemp()
|
| 52 |
-
|
| 53 |
audio = AudioSegment.from_mp3(mp3_path)
|
| 54 |
duration_ms = len(audio)
|
| 55 |
-
|
| 56 |
if duration_ms == 0:
|
|
|
|
| 57 |
return None, "فایل صوتی خالی یا بدون صدا."
|
| 58 |
-
|
| 59 |
step_size = chunk_length_ms - overlap_ms
|
| 60 |
if step_size <= 0:
|
| 61 |
step_size = chunk_length_ms // 2
|
| 62 |
num_chunks = max(1, (duration_ms // step_size) + 1)
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
|
|
|
|
|
|
| 66 |
i = 0
|
| 67 |
chunk_idx = 1
|
| 68 |
while i < duration_ms:
|
| 69 |
end_pos = min(i + chunk_length_ms, duration_ms)
|
| 70 |
chunk = audio[i:end_pos]
|
| 71 |
-
|
| 72 |
-
if len(chunk) < 3000:
|
|
|
|
| 73 |
break
|
| 74 |
-
|
| 75 |
temp_wav = os.path.join(temp_wav_dir, f"chunk_{i}.wav")
|
| 76 |
-
|
| 77 |
try:
|
| 78 |
chunk.export(temp_wav, format="wav")
|
| 79 |
-
|
|
|
|
| 80 |
text_chunk = None
|
| 81 |
retry_count = 0
|
| 82 |
max_retries = 3
|
|
|
|
| 83 |
while retry_count < max_retries:
|
| 84 |
try:
|
| 85 |
-
progress(0.5 + (i / duration_ms) * 0.5,
|
| 86 |
-
|
| 87 |
-
|
| 88 |
with sr.AudioFile(temp_wav) as source:
|
| 89 |
recognizer.adjust_for_ambient_noise(source, duration=0.5)
|
| 90 |
audio_data = recognizer.record(source, duration=None)
|
| 91 |
-
|
| 92 |
text = recognizer.recognize_google(audio_data, language=language)
|
| 93 |
if text.strip():
|
| 94 |
text_chunk = text
|
|
|
|
| 95 |
break
|
| 96 |
else:
|
| 97 |
text_chunk = "[سکوت]"
|
|
|
|
| 98 |
break
|
| 99 |
-
|
| 100 |
except sr.UnknownValueError:
|
| 101 |
text_chunk = "[نامشخص]"
|
|
|
|
| 102 |
break
|
| 103 |
-
|
| 104 |
except sr.RequestError as e:
|
| 105 |
retry_count += 1
|
| 106 |
-
|
| 107 |
-
if retry_count
|
| 108 |
-
|
|
|
|
|
|
|
| 109 |
bad_chunks += 1
|
| 110 |
-
|
|
|
|
|
|
|
| 111 |
if text_chunk:
|
| 112 |
full_text.append(text_chunk)
|
| 113 |
if "[نامشخص" in text_chunk or "[خطا" in text_chunk:
|
| 114 |
bad_chunks += 1
|
| 115 |
-
|
| 116 |
total_chunks += 1
|
| 117 |
-
|
| 118 |
except Exception as chunk_e:
|
| 119 |
-
|
|
|
|
| 120 |
full_text.append(text_chunk)
|
| 121 |
bad_chunks += 1
|
| 122 |
total_chunks += 1
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
i += step_size
|
| 129 |
chunk_idx += 1
|
| 130 |
-
|
| 131 |
shutil.rmtree(temp_wav_dir, ignore_errors=True)
|
| 132 |
-
|
| 133 |
final_text = ' '.join(full_text).strip()
|
|
|
|
|
|
|
| 134 |
if not final_text:
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
|
|
|
| 138 |
if bad_ratio > 0.7:
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
'key': 'FFmpegExtractAudio',
|
| 156 |
-
'preferredcodec': 'mp3',
|
| 157 |
-
'preferredquality': '192',
|
| 158 |
-
}],
|
| 159 |
-
'outtmpl': 'temp.%(ext)s',
|
| 160 |
-
'quiet': True,
|
| 161 |
}
|
| 162 |
-
try:
|
| 163 |
-
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
| 164 |
-
ydl.download([video_url])
|
| 165 |
-
except Exception as e:
|
| 166 |
-
return None, None, f"خطا در دانلود: {e}"
|
| 167 |
|
| 168 |
-
|
| 169 |
-
mp3_file = None
|
| 170 |
-
for f in os.listdir('.'):
|
| 171 |
-
if f.startswith('temp.') and f.endswith('.mp3'):
|
| 172 |
-
mp3_file = f
|
| 173 |
-
break
|
| 174 |
-
if not mp3_file:
|
| 175 |
-
return None, None, "فایل mp3 پیدا نشد."
|
| 176 |
-
|
| 177 |
-
# ۳. copy to temp
|
| 178 |
-
temp_mp3 = tempfile.NamedTemporaryFile(suffix='.mp3', delete=False)
|
| 179 |
-
temp_mp3.close()
|
| 180 |
-
shutil.copy2(mp3_file, temp_mp3.name)
|
| 181 |
-
os.remove(mp3_file)
|
| 182 |
-
|
| 183 |
-
progress(0.5, desc="MP3 آماده… شروع استخراج متن")
|
| 184 |
-
text, error_msg = transcribe_audio(temp_mp3.name, progress, language)
|
| 185 |
-
|
| 186 |
-
if text is None:
|
| 187 |
-
return temp_mp3.name, None, f"MP3 آماده است، ولی استخراج متن شکست خورد: {error_msg}"
|
| 188 |
-
|
| 189 |
-
progress(1.0, desc="استخراج متن کامل شد")
|
| 190 |
-
return temp_mp3.name, text, f"موفق! زبان: {language_display(language)}. {len(text.split())} کلمه استخراج شد."
|
| 191 |
-
|
| 192 |
-
|
| 193 |
-
# ------------------------------------------------------------
|
| 194 |
-
# ۳. Gradio Interface
|
| 195 |
-
# ------------------------------------------------------------
|
| 196 |
iface = gr.Interface(
|
| 197 |
fn=convert_to_mp3_and_transcribe,
|
| 198 |
inputs=[
|
| 199 |
-
gr.Textbox(
|
| 200 |
-
|
| 201 |
-
|
| 202 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 203 |
],
|
| 204 |
outputs=[
|
| 205 |
gr.File(label="دانلود MP3"),
|
| 206 |
-
gr.Textbox(label="متن استخراجشده", lines=10),
|
| 207 |
-
gr.Textbox(label="پیام وضعیت")
|
|
|
|
| 208 |
],
|
| 209 |
-
title="تبدیل ویدیو به MP3 و استخراج متن",
|
| 210 |
-
description="لینک ویدیو را وارد کنید. برای
|
| 211 |
examples=[
|
| 212 |
-
["https://www.youtube.com/watch?v=5qap5aO4i9A","fa-IR"],
|
| 213 |
-
["https://www.youtube.com/watch?v=dQw4w9WgXcQ","en-US"]
|
| 214 |
],
|
| 215 |
allow_flagging="never",
|
| 216 |
cache_examples=False
|
| 217 |
)
|
| 218 |
|
| 219 |
-
#
|
| 220 |
-
|
| 221 |
-
|
| 222 |
-
|
| 223 |
-
|
| 224 |
-
@
|
| 225 |
-
def
|
| 226 |
-
|
| 227 |
-
ورودی: {"token":"<uuid>"}
|
| 228 |
-
خروجی: {status:0/1, transcript:"..."(در صورت تکمیل)}
|
| 229 |
-
"""
|
| 230 |
-
data = request.get_json()
|
| 231 |
-
if not data or "token" not in data:
|
| 232 |
-
return jsonify({"status":0,"error":"token missing"}),400
|
| 233 |
-
|
| 234 |
-
token = data["token"]
|
| 235 |
-
with session_lock:
|
| 236 |
-
session_hash = session_map.get(token)
|
| 237 |
-
|
| 238 |
-
if not session_hash:
|
| 239 |
-
return jsonify({"status":0,"error":"unknown token"}),400
|
| 240 |
-
|
| 241 |
-
# query the gradio queue
|
| 242 |
-
import requests
|
| 243 |
-
sse_url = f"https://suprimedev-hfvd.hf.space/gradio_api/queue/data?session_hash={session_hash}"
|
| 244 |
-
resp = requests.get(sse_url, headers={
|
| 245 |
-
"Accept":"text/event-stream",
|
| 246 |
-
"Connection":"keep-alive"
|
| 247 |
-
})
|
| 248 |
-
if resp.status_code != 200:
|
| 249 |
-
return jsonify({"status":0,"error":"failed to fetch"}),500
|
| 250 |
-
|
| 251 |
-
frames = parse_sse(resp.text)
|
| 252 |
-
|
| 253 |
-
for frame in frames:
|
| 254 |
-
if frame.get('data',{}).get('msg') == 'process_completed':
|
| 255 |
-
out = frame['data'].get('output',{}).get('data')
|
| 256 |
-
if isinstance(out, list) and len(out)>=2:
|
| 257 |
-
transcript = out[1]
|
| 258 |
-
# Update DB outside – not done here (PHP handles)
|
| 259 |
-
return jsonify({"status":1,"transcript":transcript})
|
| 260 |
-
return jsonify({"status":0,"error":"not finished yet"})
|
| 261 |
|
| 262 |
-
def parse_sse(sse_text: str):
|
| 263 |
-
frames = []
|
| 264 |
-
current = {}
|
| 265 |
-
for line in sse_text.splitlines():
|
| 266 |
-
if line == '':
|
| 267 |
-
if current:
|
| 268 |
-
frames.append(current)
|
| 269 |
-
current={}
|
| 270 |
-
continue
|
| 271 |
-
if line.startswith('event:'):
|
| 272 |
-
current['event']=line[6:].strip()
|
| 273 |
-
elif line.startswith('id:'):
|
| 274 |
-
current['id']=line[3:].strip()
|
| 275 |
-
elif line.startswith('data:'):
|
| 276 |
-
current['data']=line[5:].strip()
|
| 277 |
-
if current:
|
| 278 |
-
frames.append(current)
|
| 279 |
-
|
| 280 |
-
for frame in frames:
|
| 281 |
-
if 'data' in frame:
|
| 282 |
-
try:
|
| 283 |
-
frame['data']=json.loads(frame['data'])
|
| 284 |
-
except:
|
| 285 |
-
pass
|
| 286 |
-
return frames
|
| 287 |
-
|
| 288 |
-
# ------------------------------------------------------------
|
| 289 |
-
# ۵. راهاندازی Gradio + API
|
| 290 |
-
# ------------------------------------------------------------
|
| 291 |
if __name__ == "__main__":
|
| 292 |
-
|
| 293 |
-
# 1. Gradio (بهصورت normal)
|
| 294 |
-
# 2. Flask (در پورت 5001)
|
| 295 |
-
from multiprocessing import Process
|
| 296 |
-
|
| 297 |
-
def run_gradio():
|
| 298 |
-
iface.launch(server_name="0.0.0.0", server_port=7860, share=False)
|
| 299 |
-
|
| 300 |
-
p1 = Process(target=run_gradio)
|
| 301 |
-
p1.start()
|
| 302 |
-
|
| 303 |
-
# Flask
|
| 304 |
-
app.run(host="0.0.0.0", port=5001)
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
import yt_dlp
|
| 3 |
+
import tempfile
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
import os
|
|
|
|
|
|
|
| 5 |
import shutil
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
import speech_recognition as sr
|
| 7 |
from pydub import AudioSegment
|
| 8 |
+
import time
|
| 9 |
+
import warnings
|
| 10 |
+
import json
|
| 11 |
+
from datetime import datetime, timedelta
|
| 12 |
+
import threading
|
| 13 |
+
import hashlib
|
| 14 |
|
| 15 |
warnings.filterwarnings("ignore")
|
| 16 |
|
| 17 |
+
# ذخیرهسازی نتایج برای جلوگیری از مشکل session
|
| 18 |
+
results_cache = {}
|
| 19 |
+
cache_lock = threading.Lock()
|
| 20 |
+
|
| 21 |
+
# تابع برای پاکسازی cache قدیمی (بیش از 24 ساعت)
|
| 22 |
+
def cleanup_old_cache():
|
| 23 |
+
while True:
|
| 24 |
+
time.sleep(3600) # هر ساعت چک کن
|
| 25 |
+
with cache_lock:
|
| 26 |
+
current_time = datetime.now()
|
| 27 |
+
keys_to_remove = []
|
| 28 |
+
for key, value in results_cache.items():
|
| 29 |
+
if current_time - value['timestamp'] > timedelta(hours=24):
|
| 30 |
+
keys_to_remove.append(key)
|
| 31 |
+
for key in keys_to_remove:
|
| 32 |
+
del results_cache[key]
|
| 33 |
+
if keys_to_remove:
|
| 34 |
+
print(f"[CACHE] {len(keys_to_remove)} نتیجه قدیمی پاک شد.")
|
| 35 |
+
|
| 36 |
+
# شروع thread پاکسازی
|
| 37 |
+
cleanup_thread = threading.Thread(target=cleanup_old_cache, daemon=True)
|
| 38 |
+
cleanup_thread.start()
|
| 39 |
+
|
| 40 |
+
def get_cache_key(video_url, language):
|
| 41 |
+
"""ایجاد کلید یکتا برای cache بر اساس URL و زبان"""
|
| 42 |
+
return hashlib.md5(f"{video_url}_{language}".encode()).hexdigest()
|
| 43 |
+
|
| 44 |
+
def save_result_to_cache(video_url, language, mp3_path, text, status_msg):
|
| 45 |
+
"""ذخیره نتیجه در cache"""
|
| 46 |
+
cache_key = get_cache_key(video_url, language)
|
| 47 |
+
|
| 48 |
+
# کپی فایل MP3 به مکان دائمی
|
| 49 |
+
if mp3_path and os.path.exists(mp3_path):
|
| 50 |
+
cache_mp3_path = f"cache_{cache_key}.mp3"
|
| 51 |
+
shutil.copy2(mp3_path, cache_mp3_path)
|
| 52 |
+
else:
|
| 53 |
+
cache_mp3_path = None
|
| 54 |
+
|
| 55 |
+
with cache_lock:
|
| 56 |
+
results_cache[cache_key] = {
|
| 57 |
+
'timestamp': datetime.now(),
|
| 58 |
+
'mp3_path': cache_mp3_path,
|
| 59 |
+
'text': text,
|
| 60 |
+
'status_msg': status_msg,
|
| 61 |
+
'video_url': video_url,
|
| 62 |
+
'language': language
|
| 63 |
+
}
|
| 64 |
+
|
| 65 |
+
print(f"[CACHE] نتیجه ذخیره شد: {cache_key}")
|
| 66 |
+
return cache_key
|
| 67 |
+
|
| 68 |
+
def get_result_from_cache(cache_key):
|
| 69 |
+
"""دریافت نتیجه از cache"""
|
| 70 |
+
with cache_lock:
|
| 71 |
+
if cache_key in results_cache:
|
| 72 |
+
result = results_cache[cache_key]
|
| 73 |
+
print(f"[CACHE] نتیجه یافت شد: {cache_key}")
|
| 74 |
+
return result['mp3_path'], result['text'], result['status_msg']
|
| 75 |
+
return None, None, None
|
| 76 |
+
|
| 77 |
+
def convert_to_mp3_and_transcribe(video_url, language, use_cache=True):
|
| 78 |
+
"""
|
| 79 |
+
دانلود ویدیو/صوت از لینک، تبدیل به MP3، و استخراج متن با Google Speech Recognition.
|
| 80 |
+
"""
|
| 81 |
+
if not video_url:
|
| 82 |
+
return None, None, "لینک ویدیو را وارد کنید.", None
|
| 83 |
+
|
| 84 |
+
# چک کردن cache
|
| 85 |
+
cache_key = get_cache_key(video_url, language)
|
| 86 |
+
if use_cache:
|
| 87 |
+
cached_mp3, cached_text, cached_status = get_result_from_cache(cache_key)
|
| 88 |
+
if cached_mp3 is not None or cached_text is not None:
|
| 89 |
+
return cached_mp3, cached_text, f"[از حافظه] {cached_status}", cache_key
|
| 90 |
+
|
| 91 |
+
progress = gr.Progress(track_tqdm=False)
|
| 92 |
+
|
| 93 |
+
try:
|
| 94 |
+
print(f"[DEBUG] شروع پردازش لینک: {video_url} (زبان: {language})")
|
| 95 |
+
|
| 96 |
+
progress(0, desc="شروع دانلود...")
|
| 97 |
+
|
| 98 |
+
# دانلود و تبدیل به MP3
|
| 99 |
+
ydl_opts = {
|
| 100 |
+
'format': 'bestaudio[ext=m4a]/bestaudio/best',
|
| 101 |
+
'postprocessors': [{
|
| 102 |
+
'key': 'FFmpegExtractAudio',
|
| 103 |
+
'preferredcodec': 'mp3',
|
| 104 |
+
'preferredquality': '192',
|
| 105 |
+
}],
|
| 106 |
+
'outtmpl': 'temp.%(ext)s',
|
| 107 |
+
'quiet': True,
|
| 108 |
+
}
|
| 109 |
+
|
| 110 |
+
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
| 111 |
+
ydl.download([video_url])
|
| 112 |
+
|
| 113 |
+
print("[DEBUG] دانلود کامل شد.")
|
| 114 |
+
|
| 115 |
+
# پیدا کردن MP3
|
| 116 |
+
mp3_file = None
|
| 117 |
+
for file in os.listdir('.'):
|
| 118 |
+
if file.startswith('temp.') and file.endswith('.mp3'):
|
| 119 |
+
mp3_file = file
|
| 120 |
+
break
|
| 121 |
+
|
| 122 |
+
if not mp3_file:
|
| 123 |
+
print("[DEBUG] هیچ MP3 پیدا نشد.")
|
| 124 |
+
status_msg = "خطا در دانلود یا تبدیل. مطمئن شوید لینک معتبر است."
|
| 125 |
+
save_result_to_cache(video_url, language, None, None, status_msg)
|
| 126 |
+
return None, None, status_msg, cache_key
|
| 127 |
+
|
| 128 |
+
progress(0.3, desc="دانلود کامل. کپی MP3...")
|
| 129 |
+
|
| 130 |
+
# کپی به موقت
|
| 131 |
+
temp_mp3 = tempfile.NamedTemporaryFile(suffix='.mp3', delete=False)
|
| 132 |
+
temp_mp3.close()
|
| 133 |
+
shutil.copy2(mp3_file, temp_mp3.name)
|
| 134 |
+
if os.path.exists(mp3_file):
|
| 135 |
+
os.remove(mp3_file)
|
| 136 |
+
|
| 137 |
+
print(f"[DEBUG] MP3 آماده: {temp_mp3.name}, اندازه: {os.path.getsize(temp_mp3.name)/1024:.1f} KB")
|
| 138 |
+
|
| 139 |
+
progress(0.5, desc="MP3 آماده. شروع استخراج متن با Google...")
|
| 140 |
+
|
| 141 |
+
# استخراج متن
|
| 142 |
+
text, error_msg = transcribe_audio(temp_mp3.name, progress, language)
|
| 143 |
+
|
| 144 |
+
if text is None:
|
| 145 |
+
status_msg = f"MP3 آماده است، اما استخراج متن fail شد. {error_msg}"
|
| 146 |
+
save_result_to_cache(video_url, language, temp_mp3.name, None, status_msg)
|
| 147 |
+
return temp_mp3.name, None, status_msg, cache_key
|
| 148 |
+
|
| 149 |
+
progress(1.0, desc="استخراج متن کامل شد!")
|
| 150 |
+
|
| 151 |
+
print(f"[DEBUG] متن نهایی (اولین 100 کاراکتر): {text[:100]}...")
|
| 152 |
+
|
| 153 |
+
status_msg = f"موفق! زبان: {language_display(language)}. {len(text.split())} کلمه استخراج شد."
|
| 154 |
+
save_result_to_cache(video_url, language, temp_mp3.name, text, status_msg)
|
| 155 |
+
|
| 156 |
+
return temp_mp3.name, text, status_msg, cache_key
|
| 157 |
+
|
| 158 |
+
except Exception as e:
|
| 159 |
+
print(f"[ERROR] خطای کلی: {str(e)}")
|
| 160 |
+
status_msg = f"خطا کلی: {str(e)}"
|
| 161 |
+
save_result_to_cache(video_url, language, None, None, status_msg)
|
| 162 |
+
return None, None, status_msg, cache_key
|
| 163 |
+
|
| 164 |
+
def language_display(lang_code):
|
| 165 |
+
"""نمایش نام زبان"""
|
| 166 |
+
if lang_code == 'fa-IR':
|
| 167 |
+
return "پارسی"
|
| 168 |
+
elif lang_code == 'en-US':
|
| 169 |
+
return "انگلیسی"
|
| 170 |
+
else:
|
| 171 |
+
return lang_code
|
| 172 |
+
|
| 173 |
+
def transcribe_audio(mp3_path, progress, language, chunk_length_ms=60000, overlap_ms=5000):
|
| 174 |
+
"""
|
| 175 |
+
استخراج متن با Google STT + retry برای rate limit. chunk 60s.
|
| 176 |
+
"""
|
| 177 |
recognizer = sr.Recognizer()
|
| 178 |
recognizer.energy_threshold = 300
|
| 179 |
recognizer.dynamic_energy_threshold = True
|
| 180 |
recognizer.pause_threshold = 0.8
|
| 181 |
+
|
| 182 |
full_text = []
|
| 183 |
bad_chunks = 0
|
| 184 |
total_chunks = 0
|
| 185 |
temp_wav_dir = tempfile.mkdtemp()
|
|
|
|
| 186 |
audio = AudioSegment.from_mp3(mp3_path)
|
| 187 |
duration_ms = len(audio)
|
|
|
|
| 188 |
if duration_ms == 0:
|
| 189 |
+
print("[DEBUG] فایل صوتی خالی!")
|
| 190 |
return None, "فایل صوتی خالی یا بدون صدا."
|
| 191 |
+
|
| 192 |
step_size = chunk_length_ms - overlap_ms
|
| 193 |
if step_size <= 0:
|
| 194 |
step_size = chunk_length_ms // 2
|
| 195 |
num_chunks = max(1, (duration_ms // step_size) + 1)
|
| 196 |
+
|
| 197 |
+
print(f"[DEBUG] مدت: {duration_ms/1000:.1f}s, chunkها: {num_chunks}, گام: {step_size/1000:.1f}s")
|
| 198 |
+
|
| 199 |
+
progress(0.5, desc=f"تقسیم به {num_chunks} chunk 60s (زبان: {language_display(language)})...")
|
| 200 |
+
|
| 201 |
i = 0
|
| 202 |
chunk_idx = 1
|
| 203 |
while i < duration_ms:
|
| 204 |
end_pos = min(i + chunk_length_ms, duration_ms)
|
| 205 |
chunk = audio[i:end_pos]
|
| 206 |
+
|
| 207 |
+
if len(chunk) < 3000:
|
| 208 |
+
print(f"[DEBUG] Chunk {chunk_idx} خیلی کوتاه ({len(chunk)/1000}s), رد شد.")
|
| 209 |
break
|
| 210 |
+
|
| 211 |
temp_wav = os.path.join(temp_wav_dir, f"chunk_{i}.wav")
|
| 212 |
+
|
| 213 |
try:
|
| 214 |
chunk.export(temp_wav, format="wav")
|
| 215 |
+
print(f"[DEBUG] Chunk {chunk_idx} export شد: {temp_wav}")
|
| 216 |
+
|
| 217 |
text_chunk = None
|
| 218 |
retry_count = 0
|
| 219 |
max_retries = 3
|
| 220 |
+
|
| 221 |
while retry_count < max_retries:
|
| 222 |
try:
|
| 223 |
+
progress(0.5 + (i / duration_ms) * 0.5, desc=f"Chunk {chunk_idx}/{num_chunks} ({(i/1000):.0f}-{end_pos/1000:.0f}s, retry {retry_count+1})...")
|
| 224 |
+
|
|
|
|
| 225 |
with sr.AudioFile(temp_wav) as source:
|
| 226 |
recognizer.adjust_for_ambient_noise(source, duration=0.5)
|
| 227 |
audio_data = recognizer.record(source, duration=None)
|
| 228 |
+
|
| 229 |
text = recognizer.recognize_google(audio_data, language=language)
|
| 230 |
if text.strip():
|
| 231 |
text_chunk = text
|
| 232 |
+
print(f"[DEBUG] Chunk {chunk_idx} موفق: {text[:50]}...")
|
| 233 |
break
|
| 234 |
else:
|
| 235 |
text_chunk = "[سکوت]"
|
| 236 |
+
print(f"[DEBUG] Chunk {chunk_idx} سکوت.")
|
| 237 |
break
|
| 238 |
+
|
| 239 |
except sr.UnknownValueError:
|
| 240 |
text_chunk = "[نامشخص]"
|
| 241 |
+
print(f"[DEBUG] Chunk {chunk_idx} نامشخص (نویز/سکوت).")
|
| 242 |
break
|
| 243 |
+
|
| 244 |
except sr.RequestError as e:
|
| 245 |
retry_count += 1
|
| 246 |
+
print(f"[DEBUG] Chunk {chunk_idx} RequestError (rate limit?): {str(e)}. Retry {retry_count}/{max_retries}")
|
| 247 |
+
if retry_count < max_retries:
|
| 248 |
+
time.sleep(2)
|
| 249 |
+
else:
|
| 250 |
+
text_chunk = f"[خطا rate limit: {str(e)[:30]}...]"
|
| 251 |
bad_chunks += 1
|
| 252 |
+
print(f"[DEBUG] Chunk {chunk_idx} fail پس از retryها.")
|
| 253 |
+
break
|
| 254 |
+
|
| 255 |
if text_chunk:
|
| 256 |
full_text.append(text_chunk)
|
| 257 |
if "[نامشخص" in text_chunk or "[خطا" in text_chunk:
|
| 258 |
bad_chunks += 1
|
| 259 |
+
|
| 260 |
total_chunks += 1
|
| 261 |
+
|
| 262 |
except Exception as chunk_e:
|
| 263 |
+
print(f"[ERROR] Chunk {chunk_idx} (خطای کلی): {str(chunk_e)}")
|
| 264 |
+
text_chunk = f"[خطا chunk: {str(chunk_e)[:30]}...]"
|
| 265 |
full_text.append(text_chunk)
|
| 266 |
bad_chunks += 1
|
| 267 |
total_chunks += 1
|
| 268 |
+
|
| 269 |
+
if os.path.exists(temp_wav):
|
| 270 |
+
os.remove(temp_wav)
|
| 271 |
+
|
|
|
|
| 272 |
i += step_size
|
| 273 |
chunk_idx += 1
|
| 274 |
+
|
| 275 |
shutil.rmtree(temp_wav_dir, ignore_errors=True)
|
| 276 |
+
|
| 277 |
final_text = ' '.join(full_text).strip()
|
| 278 |
+
error_msg = ""
|
| 279 |
+
|
| 280 |
if not final_text:
|
| 281 |
+
error_msg = "هیچ chunk موفقی نبود."
|
| 282 |
+
return None, error_msg
|
| 283 |
+
|
| 284 |
+
bad_ratio = bad_chunks / total_chunks if total_chunks > 0 else 1
|
| 285 |
if bad_ratio > 0.7:
|
| 286 |
+
error_msg = f"بیش از 70% chunkها fail ({bad_ratio*100:.0f}%). ممکن است نویز باشد، rate limit گوگل، یا زبان اشتباه انتخاب شده."
|
| 287 |
+
return None, error_msg
|
| 288 |
+
|
| 289 |
+
print(f"[DEBUG] {total_chunks} chunk پردازش شد, {bad_chunks} بد.")
|
| 290 |
+
|
| 291 |
+
return final_text, error_msg
|
| 292 |
+
|
| 293 |
+
# API endpoint برای دریافت نتیجه با cache key
|
| 294 |
+
def get_cached_result(cache_key):
|
| 295 |
+
"""API برای دریافت نتیجه ذخیره شده"""
|
| 296 |
+
mp3, text, status = get_result_from_cache(cache_key)
|
| 297 |
+
return {
|
| 298 |
+
"cache_key": cache_key,
|
| 299 |
+
"mp3_available": mp3 is not None,
|
| 300 |
+
"text": text,
|
| 301 |
+
"status": status
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 302 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 303 |
|
| 304 |
+
# رابط Gradio با خروجی cache key
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 305 |
iface = gr.Interface(
|
| 306 |
fn=convert_to_mp3_and_transcribe,
|
| 307 |
inputs=[
|
| 308 |
+
gr.Textbox(
|
| 309 |
+
label="لینک ویدیو (یوتیوب یا MP4 مستقیم)",
|
| 310 |
+
placeholder="https://www.youtube.com/watch?v=... یا https://example.com/video.mp4"
|
| 311 |
+
),
|
| 312 |
+
gr.Dropdown(
|
| 313 |
+
choices=[
|
| 314 |
+
("پارسی", "fa-IR"),
|
| 315 |
+
("انگلیسی", "en-US"),
|
| 316 |
+
],
|
| 317 |
+
value="fa-IR",
|
| 318 |
+
label="زبان متن"
|
| 319 |
+
),
|
| 320 |
+
gr.Checkbox(
|
| 321 |
+
label="استفاده از حافظه cache",
|
| 322 |
+
value=True,
|
| 323 |
+
visible=False # مخفی کنیم چون همیشه فعال است
|
| 324 |
+
)
|
| 325 |
],
|
| 326 |
outputs=[
|
| 327 |
gr.File(label="دانلود MP3"),
|
| 328 |
+
gr.Textbox(label="متن استخراجشده (Google STT)", lines=10),
|
| 329 |
+
gr.Textbox(label="پیام وضعیت"),
|
| 330 |
+
gr.Textbox(label="Cache Key (برای دریافت مجدد نتیجه)", visible=True)
|
| 331 |
],
|
| 332 |
+
title="تبدیل ویدیو به MP3 و استخراج متن (Google STT) - با Cache",
|
| 333 |
+
description="لینک ویدیو را وارد کنید و زبان را انتخاب کنید. نتایج برای 24 ساعت ذخیره میشوند.",
|
| 334 |
examples=[
|
| 335 |
+
["https://www.youtube.com/watch?v=5qap5aO4i9A", "fa-IR"],
|
| 336 |
+
["https://www.youtube.com/watch?v=dQw4w9WgXcQ", "en-US"]
|
| 337 |
],
|
| 338 |
allow_flagging="never",
|
| 339 |
cache_examples=False
|
| 340 |
)
|
| 341 |
|
| 342 |
+
# اضافه کردن API route برای دسترسی به cache
|
| 343 |
+
with gr.Blocks() as demo:
|
| 344 |
+
iface.render()
|
| 345 |
+
|
| 346 |
+
# API endpoint مخفی
|
| 347 |
+
@gr.route("/api/get_result/{cache_key}")
|
| 348 |
+
def api_get_result(cache_key):
|
| 349 |
+
return get_cached_result(cache_key)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 350 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 351 |
if __name__ == "__main__":
|
| 352 |
+
demo.launch(server_name="0.0.0.0", server_port=7860, share=False)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|