Update app.py

app.py

@@ -26,7 +26,27 @@ from pathlib import Path
 import gradio as gr
 import yt_dlp
 import google.generativeai as genai
+# ───────── transcript_utils.py ─────────
+from youtube_transcript_api import YouTubeTranscriptApi
+from datetime import timedelta
 
+def fetch_transcript(video_id, lang_pref=("ko","en")):
+    # available_transcripts()가 1.1.0부터 추가됨
+    for lang in lang_pref:
+        try:
+            transcript = YouTubeTranscriptApi.get_transcript(video_id, languages=[lang])
+            break
+        except Exception:
+            continue
+    else:
+        raise RuntimeError("자막을 찾을 수 없습니다.")
+
+    lines = []
+    for seg in transcript:
+        t = str(timedelta(seconds=int(seg["start"])))  # 0:01:23
+        t_mmss = ":".join(t.split(":")[-2:])           # 01:23
+        lines.append(f"**[{t_mmss}]** {seg['text']}")
+    return "\n".join(lines)
 # ──────────────────────────────────────────────────────────────
 # 기본 쿠키 파일 경로 ― 파일명이 동일하면 자동 사용
 # ──────────────────────────────────────────────────────────────
@@ -623,20 +643,410 @@ def download_with_cookies(url, quality, audio_only, cookies_file, progress=gr.Pr
 # =================================================================
 # Gradio UI
 # =================================================================
+#!/usr/bin/env python3
+"""
+YouTube Video Analyzer & Downloader Pro
+───────────────────────────────────────
+•  `www.youtube.com_cookies.txt` 가 app.py 와 같은 폴더에 있으면 자동으로 사용
+•  UI에서 쿠키를 업로드하면 그 파일이 *우선* 적용
+•  “Transcript” 탭을 추가해 **전체 자막 + MM:SS 타임스탬프** 출력
+"""
+
+# ── 표준 라이브러리 ───────────────────────────────────────────
+import os, re, json, shutil, tempfile
+from datetime import datetime, timedelta
+from pathlib import Path
+
+# ── 서드파티 ──────────────────────────────────────────────────
+import gradio as gr
+import yt_dlp
+import google.generativeai as genai
+from youtube_transcript_api import YouTubeTranscriptApi  # NEW
+
+# ── 상수 ──────────────────────────────────────────────────────
+DEFAULT_COOKIE_FILE = Path(__file__).with_name("www.youtube.com_cookies.txt")
+
+# YouTube URL 정규식(캡처 그룹 6이 영상 ID)
+_YT_REGEX = re.compile(
+    r"(https?://)?(www\.)?"
+    r"(youtube|youtu|youtube-nocookie)\.(com|be)/"
+    r"(watch\?v=|embed/|v/|.+\?v=)?([^&=%\?]{11})"
+)
+
+
+# =================================================================
+# Helper :  video-ID 추출 + 자막 가져오기
+# =================================================================
+def extract_video_id(url: str) -> str | None:
+    """유튜브 URL에서 11-글자 비디오 ID 반환(없으면 None)"""
+    m = _YT_REGEX.match(url)
+    return m.group(6) if m else None
+
+
+def fetch_transcript(video_id: str, pref_lang=("ko", "en")) -> str:
+    """
+    유튜브 자막을 가져와
+    **[MM:SS]** line 형식으로 결합한 뒤 문자열로 반환.
+    """
+    transcript = None
+    # 언어 우선순위대로 시도
+    for lang in pref_lang:
+        try:
+            transcript = YouTubeTranscriptApi.get_transcript(video_id, languages=[lang])
+            break
+        except Exception:
+            continue
+    # 그래도 실패하면 임의 언어
+    if transcript is None:
+        transcript = YouTubeTranscriptApi.get_transcript(video_id)
+
+    lines = []
+    for seg in transcript:
+        t = str(timedelta(seconds=int(seg["start"])))       # H:MM:SS
+        t_mmss = ":".join(t.split(":")[-2:])                # MM:SS
+        lines.append(f"**[{t_mmss}]** {seg['text']}")
+    return "\n".join(lines)
+
+
+# =================================================================
+# 메인 클래스
+# =================================================================
+class YouTubeDownloader:
+    def __init__(self):
+        self.download_dir = tempfile.mkdtemp()
+        self.temp_downloads = tempfile.mkdtemp(prefix="youtube_downloads_")
+        self.downloads_folder = os.path.join(
+            os.path.expanduser("~"), "Downloads", "YouTube_Downloads"
+        )
+        os.makedirs(self.downloads_folder, exist_ok=True)
+        self.gemini_model = None
+
+    # ───────── Gemini ─────────
+    def configure_gemini(self, api_key):
+        try:
+            genai.configure(api_key=api_key)
+            self.gemini_model = genai.GenerativeModel(
+                model_name="gemini-1.5-flash-latest"
+            )
+            return True, "✅ Gemini API configured successfully!"
+        except Exception as e:
+            return False, f"❌ Failed to configure Gemini API: {e}"
+
+    # ───────── 정리 ─────────
+    def cleanup(self):
+        try:
+            if os.path.exists(self.download_dir):
+                shutil.rmtree(self.download_dir)
+            if os.path.exists(self.temp_downloads):
+                shutil.rmtree(self.temp_downloads)
+        except Exception:
+            pass
+
+    # ───────── URL 검증 ──────
+    def is_valid_youtube_url(self, url):
+        return _YT_REGEX.match(url) is not None
+
+    # ───────── Gemini scene breakdown (생략 없이 전체 구현) ──────
+    def generate_scene_breakdown_gemini(self, video_info):
+        if not self.gemini_model:
+            return self.generate_scene_breakdown_fallback(video_info)
+        try:
+            duration = video_info.get("duration", 0)
+            title = video_info.get("title", "")
+            description = video_info.get("description", "")[:1500]
+            if not duration:
+                return [
+                    "**[Duration Unknown]**: Unable to generate timestamped breakdown – "
+                    "video duration not available"
+                ]
+
+            prompt = f"""
+Analyze this YouTube video and create a highly detailed, scene-by-scene breakdown
+with precise timestamps and specific descriptions:
+
+Title: {title}
+Duration: {duration} seconds
+Description: {description}
+
+IMPORTANT INSTRUCTIONS:
+1. Create detailed scene descriptions that include:
+   - Physical appearance of people (age, gender, clothing, hair, etc.)
+   - Exact actions being performed
+   - Dialogue or speech (include actual lines if audible, or infer probable spoken
+     lines based on actions and setting; format them as "Character: line…")
+   - Setting and environment details
+   - Props, objects, or products being shown
+   - Visual effects, text overlays, or graphics
+   - Mood, tone, and atmosphere
+   - Camera movements or angles (if apparent)
+2. Dialogue Emphasis:
+   - Include short dialogue lines in **every scene** wherever plausible.
+   - Write lines like: Character: "Actual or inferred line…"
+   - If dialogue is not available, intelligently infer probable phrases
+3. Timestamp Guidelines:
+   - <1 min : 2–3 s | 1–5 min : 3–5 s | 5–15 min : 5–10 s | >15 min : 10–15 s
+   - Max 20 scenes
+4. Format: **[MM:SS-MM:SS]** description
+"""
+            resp = self.gemini_model.generate_content(prompt)
+            if not resp or not resp.text:
+                return self.generate_scene_breakdown_fallback(video_info)
+
+            scenes, cur = [], ""
+            for line in resp.text.splitlines():
+                line = line.strip()
+                if line.startswith("**[") and "]**:" in line:
+                    if cur:
+                        scenes.append(cur.strip())
+                    cur = line
+                elif cur:
+                    cur += "\n" + line
+            if cur:
+                scenes.append(cur.strip())
+            return scenes if scenes else self.generate_scene_breakdown_fallback(video_info)
+        except Exception:
+            return self.generate_scene_breakdown_fallback(video_info)
+
+    # ───────── fallback breakdown ──────
+    def generate_scene_breakdown_fallback(self, video_info):
+        duration = video_info.get("duration", 0)
+        if not duration:
+            return ["**[Duration Unknown]**: Unable to generate timestamped breakdown"]
+
+        if duration <= 60:
+            seg = 3
+        elif duration <= 300:
+            seg = 5
+        elif duration <= 900:
+            seg = 10
+        else:
+            seg = 15
+
+        total = min(duration // seg + 1, 20)
+        vtype = self.detect_video_type_detailed(
+            video_info.get("title", ""), video_info.get("description", "")
+        )
+        scenes = []
+        for i in range(total):
+            s, e = i * seg, min(i * seg + seg - 1, duration)
+            scenes.append(
+                f"**[{s//60:02d}:{s%60:02d}-{e//60:02d}:{e%60:02d}]**: "
+                f"{self.generate_contextual_description(i, total, vtype, '', video_info.get('title',''))}"
+            )
+        return scenes
+
+    # ───────── detect helpers (상세) ──────
+    def detect_video_type_detailed(self, title, desc):
+        t = (title + " " + desc).lower()
+        if any(x in t for x in ["tutorial", "how to", "guide", "diy"]):
+            return "tutorial"
+        if any(x in t for x in ["review", "unboxing", "comparison"]):
+            return "review"
+        if any(x in t for x in ["vlog", "daily", "routine"]):
+            return "vlog"
+        if any(x in t for x in ["music", "song", "cover"]):
+            return "music"
+        if any(x in t for x in ["comedy", "prank", "challenge"]):
+            return "entertainment"
+        if any(x in t for x in ["news", "update", "report"]):
+            return "news"
+        if any(x in t for x in ["cooking", "recipe", "food"]):
+            return "cooking"
+        if any(x in t for x in ["workout", "fitness", "yoga"]):
+            return "fitness"
+        return "general"
+
+    def generate_contextual_description(
+        self, idx, total, vtype, uploader, title
+    ):
+        if idx == 0:
+            return "The creator greets viewers and introduces the video."
+        if idx == total - 1:
+            return "The creator wraps up and thanks viewers."
+        return "Content continues according to the video type."
+
+    # ───────── quick-detect helpers (요약) ──────
+    def detect_video_type(self, title, desc):
+        t = (title + " " + desc).lower()
+        if any(x in t for x in ["music", "song", "album"]):
+            return "🎵 Music"
+        if any(x in t for x in ["tutorial", "guide"]):
+            return "📚 Tutorial"
+        if any(x in t for x in ["comedy", "vlog"]):
+            return "🎭 Entertainment"
+        if any(x in t for x in ["news", "report"]):
+            return "📰 News"
+        if any(x in t for x in ["review", "unboxing"]):
+            return "⭐ Review"
+        return "🎬 General"
+
+    def detect_background_music(self, video_info):
+        title = video_info.get("title", "").lower()
+        if "music" in title:
+            return "🎵 Original music"
+        if "tutorial" in title:
+            return "🔇 Minimal music"
+        return "🎼 Background music"
+
+    def detect_influencer_status(self, video_info):
+        subs = video_info.get("channel_followers", 0)
+        if subs > 10_000_000:
+            return "🌟 Mega (10 M+)"
+        if subs > 1_000_000:
+            return "⭐ Major (1 M+)"
+        if subs > 100_000:
+            return "🎯 Mid (100 K+)"
+        return "👤"
+
+    @staticmethod
+    def format_number(n):
+        if n >= 1_000_000:
+            return f"{n/1_000_000:.1f} M"
+        if n >= 1_000:
+            return f"{n/1_000:.1f} K"
+        return str(n)
+
+    # ───────── 리포트 ──────
+    def format_video_info(self, info):
+        title = info.get("title", "")
+        uploader = info.get("uploader", "")
+        duration = info.get("duration", 0)
+        dur = f"{duration//60}:{duration%60:02d}"
+        views = info.get("view_count", 0)
+        likes = info.get("like_count", 0)
+        comments = info.get("comment_count", 0)
+        scenes = self.generate_scene_breakdown_gemini(info)
+
+        return f"""
+🎬 **{title}**
+Uploader: {uploader} Duration: {dur}
+
+Views / Likes / Comments: {self.format_number(views)} / {self.format_number(likes)} / {self.format_number(comments)}
+
+{'-'*48}
+{"".join(scenes)}
+"""
+
+    # ───────── 메타데이터 추출 ──────
+    def get_video_info(self, url, progress=gr.Progress(), cookiefile=None):
+        if not self.is_valid_youtube_url(url):
+            return None, "❌ Invalid URL"
+
+        if cookiefile and os.path.exists(cookiefile):
+            cookiefile = cookiefile
+        elif DEFAULT_COOKIE_FILE.exists():
+            cookiefile = str(DEFAULT_COOKIE_FILE)
+        else:
+            cookiefile = None
+
+        try:
+            ydl_opts = {"noplaylist": True, "quiet": True}
+            if cookiefile:
+                ydl_opts["cookiefile"] = cookiefile
+            with yt_dlp.YoutubeDL(ydl_opts) as ydl:
+                info = ydl.extract_info(url, download=False)
+            return info, "OK"
+        except Exception as e:
+            return None, f"yt-dlp error: {e}"
+
+    # ───────── 다운로드 ──────
+    def download_video(
+        self, url, quality="best", audio_only=False, progress=gr.Progress(), cookiefile=None
+    ):
+        if not self.is_valid_youtube_url(url):
+            return None, "❌ Invalid URL"
+        if cookiefile and os.path.exists(cookiefile):
+            cookiefile = cookiefile
+        elif DEFAULT_COOKIE_FILE.exists():
+            cookiefile = str(DEFAULT_COOKIE_FILE)
+        else:
+            cookiefile = None
+
+        ts = datetime.now().strftime("%Y%m%d_%H%M%S")
+        ydl_opts = {
+            "outtmpl": os.path.join(self.temp_downloads, f"%(title)s_{ts}.%(ext)s"),
+            "noplaylist": True,
+        }
+        if audio_only:
+            ydl_opts["format"] = "bestaudio/best"
+            ydl_opts["postprocessors"] = [
+                {"key": "FFmpegExtractAudio", "preferredcodec": "mp3", "preferredquality": "192"}
+            ]
+        else:
+            if quality == "720p":
+                ydl_opts["format"] = "best[height<=720]"
+            elif quality == "480p":
+                ydl_opts["format"] = "best[height<=480]"
+            else:
+                ydl_opts["format"] = "best[height<=1080]"
+        if cookiefile:
+            ydl_opts["cookiefile"] = cookiefile
+
+        try:
+            with yt_dlp.YoutubeDL(ydl_opts) as ydl:
+                ydl.extract_info(url, download=True)
+            # 첫 파일 찾기
+            for f in os.listdir(self.temp_downloads):
+                if ts in f:
+                    temp_fp = os.path.join(self.temp_downloads, f)
+                    final_fp = os.path.join(self.downloads_folder, f)
+                    try:
+                        shutil.copy2(temp_fp, final_fp)
+                        saved = final_fp
+                    except Exception:
+                        saved = temp_fp
+                    return temp_fp, f"✅ Saved: {saved}"
+            return None, "❌ Downloaded file not found"
+        except Exception as e:
+            return None, f"❌ Download failed: {e}"
+
+
+# =================================================================
+# Gradio Helper 함수
+# =================================================================
+downloader = YouTubeDownloader()
+
+
+def configure_api_key(api_key):
+    ok, msg = downloader.configure_gemini(api_key.strip()) if api_key else (False, "❌ API key required")
+    return msg, gr.update(visible=ok)
+
+
+def analyze_with_cookies(url, cookies_file, progress=gr.Progress()):
+    info, err = downloader.get_video_info(url, progress, cookies_file)
+    return downloader.format_video_info(info) if info else f"❌ {err}"
+
+
+def download_with_cookies(url, qual, audio, cookies_file, progress=gr.Progress()):
+    fp, msg = downloader.download_video(url, qual, audio, progress, cookies_file)
+    return fp, msg
+
+
+def get_transcript(url, cookies_file):
+    vid = extract_video_id(url)
+    if not vid:
+        return "❌ Invalid YouTube URL"
+    try:
+        return fetch_transcript(vid)
+    except Exception as e:
+        return f"❌ {e}"
+
+
+# =================================================================
+# UI
+# =================================================================
 def create_interface():
     with gr.Blocks(
         theme=gr.themes.Soft(), title="🎥 YouTube Video Analyzer & Downloader Pro"
     ) as iface:
         gr.HTML("<h1>🎥 YouTube Video Analyzer & Downloader Pro</h1>")
 
-        # API 섹션
+        # API 설정
         with gr.Group():
             gr.HTML("<h3>🔑 Google Gemini API Configuration</h3>")
             with gr.Row():
                 api_key_in = gr.Textbox(
-                    label="🔑 Google API Key",
-                    placeholder="Paste your Google API key…",
-                    type="password",
+                    label="🔑 Google API Key", type="password", placeholder="Paste your Google API key…"
                 )
                 api_btn = gr.Button("🔧 Configure API", variant="secondary")
             api_status = gr.Textbox(
@@ -646,71 +1056,550 @@ def create_interface():
                 lines=1,
             )
 
-        # 메인 UI
+        # 공통 입력
         with gr.Row():
-            url_in = gr.Textbox(
-                label="🔗 YouTube URL",
-                placeholder="Paste YouTube video URL…",
-            )
+            url_in = gr.Textbox(label="🔗 YouTube URL", placeholder="Paste YouTube video URL…")
             cookies_in = gr.File(
-                label="🍪 Upload cookies.txt (optional)",
-                file_types=[".txt"],
-                type="filepath",
+                label="🍪 Upload cookies.txt (optional)", file_types=[".txt"], type="filepath"
             )
 
         with gr.Tabs():
+            # 분석 탭
             with gr.TabItem("📊 Video Analysis"):
                 analyze_btn = gr.Button("🔍 Analyze Video", variant="primary")
-                analysis_out = gr.Textbox(
-                    label="📊 Analysis Report", lines=25, show_copy_button=True
-                )
+                analysis_out = gr.Textbox(label="📊 Analysis Report", lines=30, show_copy_button=True)
                 analyze_btn.click(
-                    fn=analyze_with_cookies,
-                    inputs=[url_in, cookies_in],
-                    outputs=analysis_out,
+                    analyze_with_cookies, inputs=[url_in, cookies_in], outputs=analysis_out, show_progress=True
+                )
+            # 다운로드 탭
+            with gr.TabItem("⬇️ Video Download"):
+                with gr.Row():
+                    quality_dd = gr.Dropdown(
+                        choices=["best", "720p", "480p"], value="best", label="📺 Quality"
+                    )
+                    audio_cb = gr.Checkbox(label="🎵 Audio only (MP3)")
+                download_btn = gr.Button("⬇️ Download Video", variant="primary")
+                dl_status = gr.Textbox(label="📥 Download Status", lines=5, show_copy_button=True)
+                dl_file = gr.File(label="📁 Downloaded File", visible=False)
+
+                def wrapped_dl(u, q, a, c, prog=gr.Progress()):
+                    fp, st = download_with_cookies(u, q, a, c, prog)
+                    return (st, gr.update(value=fp, visible=True)) if fp and os.path.exists(fp) else (
+                        st,
+                        gr.update(visible=False),
+                    )
+
+                download_btn.click(
+                    wrapped_dl,
+                    inputs=[url_in, quality_dd, audio_cb, cookies_in],
+                    outputs=[dl_status, dl_file],
                     show_progress=True,
                 )
+            # 자막 탭  NEW
+            with gr.TabItem("🗒️ Transcript"):
+                tr_btn = gr.Button("📜 Get Full Transcript", variant="primary")
+                tr_out = gr.Textbox(
+                    label="🗒️ Transcript (full)", lines=30, show_copy_button=True
+                )
+                tr_btn.click(
+                    get_transcript, inputs=[url_in, cookies_in], outputs=tr_out, show_progress=True
+                )
+
+        # API 버튼
+        api_btn.click(configure_api_key, inputs=[api_key_in], outputs=[api_status])
+
+        gr.HTML(
+            """
+<div style="margin-top:20px;padding:15px;background:#f0f8ff;border-left:5px solid #4285f4;border-radius:10px;">
+  <h3>💡 Tip: 쿠키 파일 자동 사용</h3>
+  <p><code>www.youtube.com_cookies.txt</code> 파일을 <strong>app.py</strong>와 같은
+  폴더에 두면 업로드 없이 자동 사용됩니다.</p>
+</div>
+"""
+        )
+    return iface
+
+
+# =================================================================
+# 실행
+# =================================================================
+if __name__ == "__main__":
+    demo = create_interface()
+    import atexit
+
+    atexit.register(downloader.cleanup)
+    demo.launch(debug=True, show_error=True)
+#!/usr/bin/env python3
+"""
+YouTube Video Analyzer & Downloader Pro
+───────────────────────────────────────
+•  `www.youtube.com_cookies.txt` 가 app.py 와 같은 폴더에 있으면 자동으로 사용
+•  UI에서 쿠키를 업로드하면 그 파일이 *우선* 적용
+•  “Transcript” 탭을 추가해 **전체 자막 + MM:SS 타임스탬프** 출력
+"""
+
+# ── 표준 라이브러리 ───────────────────────────────────────────
+import os, re, json, shutil, tempfile
+from datetime import datetime, timedelta
+from pathlib import Path
+
+# ── 서드파티 ──────────────────────────────────────────────────
+import gradio as gr
+import yt_dlp
+import google.generativeai as genai
+from youtube_transcript_api import YouTubeTranscriptApi  # NEW
+
+# ── 상수 ──────────────────────────────────────────────────────
+DEFAULT_COOKIE_FILE = Path(__file__).with_name("www.youtube.com_cookies.txt")
+
+# YouTube URL 정규식(캡처 그룹 6이 영상 ID)
+_YT_REGEX = re.compile(
+    r"(https?://)?(www\.)?"
+    r"(youtube|youtu|youtube-nocookie)\.(com|be)/"
+    r"(watch\?v=|embed/|v/|.+\?v=)?([^&=%\?]{11})"
+)
+
+
+# =================================================================
+# Helper :  video-ID 추출 + 자막 가져오기
+# =================================================================
+def extract_video_id(url: str) -> str | None:
+    """유튜브 URL에서 11-글자 비디오 ID 반환(없으면 None)"""
+    m = _YT_REGEX.match(url)
+    return m.group(6) if m else None
+
+
+def fetch_transcript(video_id: str, pref_lang=("ko", "en")) -> str:
+    """
+    유튜브 자막을 가져와
+    **[MM:SS]** line 형식으로 결합한 뒤 문자열로 반환.
+    """
+    transcript = None
+    # 언어 우선순위대로 시도
+    for lang in pref_lang:
+        try:
+            transcript = YouTubeTranscriptApi.get_transcript(video_id, languages=[lang])
+            break
+        except Exception:
+            continue
+    # 그래도 실패하면 임의 언어
+    if transcript is None:
+        transcript = YouTubeTranscriptApi.get_transcript(video_id)
+
+    lines = []
+    for seg in transcript:
+        t = str(timedelta(seconds=int(seg["start"])))       # H:MM:SS
+        t_mmss = ":".join(t.split(":")[-2:])                # MM:SS
+        lines.append(f"**[{t_mmss}]** {seg['text']}")
+    return "\n".join(lines)
+
+
+# =================================================================
+# 메인 클래스
+# =================================================================
+class YouTubeDownloader:
+    def __init__(self):
+        self.download_dir = tempfile.mkdtemp()
+        self.temp_downloads = tempfile.mkdtemp(prefix="youtube_downloads_")
+        self.downloads_folder = os.path.join(
+            os.path.expanduser("~"), "Downloads", "YouTube_Downloads"
+        )
+        os.makedirs(self.downloads_folder, exist_ok=True)
+        self.gemini_model = None
+
+    # ───────── Gemini ─────────
+    def configure_gemini(self, api_key):
+        try:
+            genai.configure(api_key=api_key)
+            self.gemini_model = genai.GenerativeModel(
+                model_name="gemini-1.5-flash-latest"
+            )
+            return True, "✅ Gemini API configured successfully!"
+        except Exception as e:
+            return False, f"❌ Failed to configure Gemini API: {e}"
+
+    # ───────── 정리 ─────────
+    def cleanup(self):
+        try:
+            if os.path.exists(self.download_dir):
+                shutil.rmtree(self.download_dir)
+            if os.path.exists(self.temp_downloads):
+                shutil.rmtree(self.temp_downloads)
+        except Exception:
+            pass
+
+    # ───────── URL 검증 ──────
+    def is_valid_youtube_url(self, url):
+        return _YT_REGEX.match(url) is not None
+
+    # ───────── Gemini scene breakdown (생략 없이 전체 구현) ──────
+    def generate_scene_breakdown_gemini(self, video_info):
+        if not self.gemini_model:
+            return self.generate_scene_breakdown_fallback(video_info)
+        try:
+            duration = video_info.get("duration", 0)
+            title = video_info.get("title", "")
+            description = video_info.get("description", "")[:1500]
+            if not duration:
+                return [
+                    "**[Duration Unknown]**: Unable to generate timestamped breakdown – "
+                    "video duration not available"
+                ]
+
+            prompt = f"""
+Analyze this YouTube video and create a highly detailed, scene-by-scene breakdown
+with precise timestamps and specific descriptions:
+
+Title: {title}
+Duration: {duration} seconds
+Description: {description}
+
+IMPORTANT INSTRUCTIONS:
+1. Create detailed scene descriptions that include:
+   - Physical appearance of people (age, gender, clothing, hair, etc.)
+   - Exact actions being performed
+   - Dialogue or speech (include actual lines if audible, or infer probable spoken
+     lines based on actions and setting; format them as "Character: line…")
+   - Setting and environment details
+   - Props, objects, or products being shown
+   - Visual effects, text overlays, or graphics
+   - Mood, tone, and atmosphere
+   - Camera movements or angles (if apparent)
+2. Dialogue Emphasis:
+   - Include short dialogue lines in **every scene** wherever plausible.
+   - Write lines like: Character: "Actual or inferred line…"
+   - If dialogue is not available, intelligently infer probable phrases
+3. Timestamp Guidelines:
+   - <1 min : 2–3 s | 1–5 min : 3–5 s | 5–15 min : 5–10 s | >15 min : 10–15 s
+   - Max 20 scenes
+4. Format: **[MM:SS-MM:SS]** description
+"""
+            resp = self.gemini_model.generate_content(prompt)
+            if not resp or not resp.text:
+                return self.generate_scene_breakdown_fallback(video_info)
+
+            scenes, cur = [], ""
+            for line in resp.text.splitlines():
+                line = line.strip()
+                if line.startswith("**[") and "]**:" in line:
+                    if cur:
+                        scenes.append(cur.strip())
+                    cur = line
+                elif cur:
+                    cur += "\n" + line
+            if cur:
+                scenes.append(cur.strip())
+            return scenes if scenes else self.generate_scene_breakdown_fallback(video_info)
+        except Exception:
+            return self.generate_scene_breakdown_fallback(video_info)
+
+    # ───────── fallback breakdown ──────
+    def generate_scene_breakdown_fallback(self, video_info):
+        duration = video_info.get("duration", 0)
+        if not duration:
+            return ["**[Duration Unknown]**: Unable to generate timestamped breakdown"]
+
+        if duration <= 60:
+            seg = 3
+        elif duration <= 300:
+            seg = 5
+        elif duration <= 900:
+            seg = 10
+        else:
+            seg = 15
+
+        total = min(duration // seg + 1, 20)
+        vtype = self.detect_video_type_detailed(
+            video_info.get("title", ""), video_info.get("description", "")
+        )
+        scenes = []
+        for i in range(total):
+            s, e = i * seg, min(i * seg + seg - 1, duration)
+            scenes.append(
+                f"**[{s//60:02d}:{s%60:02d}-{e//60:02d}:{e%60:02d}]**: "
+                f"{self.generate_contextual_description(i, total, vtype, '', video_info.get('title',''))}"
+            )
+        return scenes
+
+    # ───────── detect helpers (상세) ──────
+    def detect_video_type_detailed(self, title, desc):
+        t = (title + " " + desc).lower()
+        if any(x in t for x in ["tutorial", "how to", "guide", "diy"]):
+            return "tutorial"
+        if any(x in t for x in ["review", "unboxing", "comparison"]):
+            return "review"
+        if any(x in t for x in ["vlog", "daily", "routine"]):
+            return "vlog"
+        if any(x in t for x in ["music", "song", "cover"]):
+            return "music"
+        if any(x in t for x in ["comedy", "prank", "challenge"]):
+            return "entertainment"
+        if any(x in t for x in ["news", "update", "report"]):
+            return "news"
+        if any(x in t for x in ["cooking", "recipe", "food"]):
+            return "cooking"
+        if any(x in t for x in ["workout", "fitness", "yoga"]):
+            return "fitness"
+        return "general"
+
+    def generate_contextual_description(
+        self, idx, total, vtype, uploader, title
+    ):
+        if idx == 0:
+            return "The creator greets viewers and introduces the video."
+        if idx == total - 1:
+            return "The creator wraps up and thanks viewers."
+        return "Content continues according to the video type."
+
+    # ───────── quick-detect helpers (요약) ──────
+    def detect_video_type(self, title, desc):
+        t = (title + " " + desc).lower()
+        if any(x in t for x in ["music", "song", "album"]):
+            return "🎵 Music"
+        if any(x in t for x in ["tutorial", "guide"]):
+            return "📚 Tutorial"
+        if any(x in t for x in ["comedy", "vlog"]):
+            return "🎭 Entertainment"
+        if any(x in t for x in ["news", "report"]):
+            return "📰 News"
+        if any(x in t for x in ["review", "unboxing"]):
+            return "⭐ Review"
+        return "🎬 General"
+
+    def detect_background_music(self, video_info):
+        title = video_info.get("title", "").lower()
+        if "music" in title:
+            return "🎵 Original music"
+        if "tutorial" in title:
+            return "🔇 Minimal music"
+        return "🎼 Background music"
+
+    def detect_influencer_status(self, video_info):
+        subs = video_info.get("channel_followers", 0)
+        if subs > 10_000_000:
+            return "🌟 Mega (10 M+)"
+        if subs > 1_000_000:
+            return "⭐ Major (1 M+)"
+        if subs > 100_000:
+            return "🎯 Mid (100 K+)"
+        return "👤"
+
+    @staticmethod
+    def format_number(n):
+        if n >= 1_000_000:
+            return f"{n/1_000_000:.1f} M"
+        if n >= 1_000:
+            return f"{n/1_000:.1f} K"
+        return str(n)
+
+    # ───────── 리포트 ──────
+    def format_video_info(self, info):
+        title = info.get("title", "")
+        uploader = info.get("uploader", "")
+        duration = info.get("duration", 0)
+        dur = f"{duration//60}:{duration%60:02d}"
+        views = info.get("view_count", 0)
+        likes = info.get("like_count", 0)
+        comments = info.get("comment_count", 0)
+        scenes = self.generate_scene_breakdown_gemini(info)
+
+        return f"""
+🎬 **{title}**
+Uploader: {uploader} Duration: {dur}
+
+Views / Likes / Comments: {self.format_number(views)} / {self.format_number(likes)} / {self.format_number(comments)}
+
+{'-'*48}
+{"".join(scenes)}
+"""
+
+    # ───────── 메타데이터 추출 ──────
+    def get_video_info(self, url, progress=gr.Progress(), cookiefile=None):
+        if not self.is_valid_youtube_url(url):
+            return None, "❌ Invalid URL"
+
+        if cookiefile and os.path.exists(cookiefile):
+            cookiefile = cookiefile
+        elif DEFAULT_COOKIE_FILE.exists():
+            cookiefile = str(DEFAULT_COOKIE_FILE)
+        else:
+            cookiefile = None
+
+        try:
+            ydl_opts = {"noplaylist": True, "quiet": True}
+            if cookiefile:
+                ydl_opts["cookiefile"] = cookiefile
+            with yt_dlp.YoutubeDL(ydl_opts) as ydl:
+                info = ydl.extract_info(url, download=False)
+            return info, "OK"
+        except Exception as e:
+            return None, f"yt-dlp error: {e}"
+
+    # ───────── 다운로드 ──────
+    def download_video(
+        self, url, quality="best", audio_only=False, progress=gr.Progress(), cookiefile=None
+    ):
+        if not self.is_valid_youtube_url(url):
+            return None, "❌ Invalid URL"
+        if cookiefile and os.path.exists(cookiefile):
+            cookiefile = cookiefile
+        elif DEFAULT_COOKIE_FILE.exists():
+            cookiefile = str(DEFAULT_COOKIE_FILE)
+        else:
+            cookiefile = None
+
+        ts = datetime.now().strftime("%Y%m%d_%H%M%S")
+        ydl_opts = {
+            "outtmpl": os.path.join(self.temp_downloads, f"%(title)s_{ts}.%(ext)s"),
+            "noplaylist": True,
+        }
+        if audio_only:
+            ydl_opts["format"] = "bestaudio/best"
+            ydl_opts["postprocessors"] = [
+                {"key": "FFmpegExtractAudio", "preferredcodec": "mp3", "preferredquality": "192"}
+            ]
+        else:
+            if quality == "720p":
+                ydl_opts["format"] = "best[height<=720]"
+            elif quality == "480p":
+                ydl_opts["format"] = "best[height<=480]"
+            else:
+                ydl_opts["format"] = "best[height<=1080]"
+        if cookiefile:
+            ydl_opts["cookiefile"] = cookiefile
+
+        try:
+            with yt_dlp.YoutubeDL(ydl_opts) as ydl:
+                ydl.extract_info(url, download=True)
+            # 첫 파일 찾기
+            for f in os.listdir(self.temp_downloads):
+                if ts in f:
+                    temp_fp = os.path.join(self.temp_downloads, f)
+                    final_fp = os.path.join(self.downloads_folder, f)
+                    try:
+                        shutil.copy2(temp_fp, final_fp)
+                        saved = final_fp
+                    except Exception:
+                        saved = temp_fp
+                    return temp_fp, f"✅ Saved: {saved}"
+            return None, "❌ Downloaded file not found"
+        except Exception as e:
+            return None, f"❌ Download failed: {e}"
+
+
+# =================================================================
+# Gradio Helper 함수
+# =================================================================
+downloader = YouTubeDownloader()
+
+
+def configure_api_key(api_key):
+    ok, msg = downloader.configure_gemini(api_key.strip()) if api_key else (False, "❌ API key required")
+    return msg, gr.update(visible=ok)
+
+
+def analyze_with_cookies(url, cookies_file, progress=gr.Progress()):
+    info, err = downloader.get_video_info(url, progress, cookies_file)
+    return downloader.format_video_info(info) if info else f"❌ {err}"
+
+
+def download_with_cookies(url, qual, audio, cookies_file, progress=gr.Progress()):
+    fp, msg = downloader.download_video(url, qual, audio, progress, cookies_file)
+    return fp, msg
+
+
+def get_transcript(url, cookies_file):
+    vid = extract_video_id(url)
+    if not vid:
+        return "❌ Invalid YouTube URL"
+    try:
+        return fetch_transcript(vid)
+    except Exception as e:
+        return f"❌ {e}"
+
+
+# =================================================================
+# UI
+# =================================================================
+def create_interface():
+    with gr.Blocks(
+        theme=gr.themes.Soft(), title="🎥 YouTube Video Analyzer & Downloader Pro"
+    ) as iface:
+        gr.HTML("<h1>🎥 YouTube Video Analyzer & Downloader Pro</h1>")
+
+        # API 설정
+        with gr.Group():
+            gr.HTML("<h3>🔑 Google Gemini API Configuration</h3>")
+            with gr.Row():
+                api_key_in = gr.Textbox(
+                    label="🔑 Google API Key", type="password", placeholder="Paste your Google API key…"
+                )
+                api_btn = gr.Button("🔧 Configure API", variant="secondary")
+            api_status = gr.Textbox(
+                label="API Status",
+                value="❌ Gemini API not configured – Using fallback analysis",
+                interactive=False,
+                lines=1,
+            )
 
+        # 공통 입력
+        with gr.Row():
+            url_in = gr.Textbox(label="🔗 YouTube URL", placeholder="Paste YouTube video URL…")
+            cookies_in = gr.File(
+                label="🍪 Upload cookies.txt (optional)", file_types=[".txt"], type="filepath"
+            )
+
+        with gr.Tabs():
+            # 분석 탭
+            with gr.TabItem("📊 Video Analysis"):
+                analyze_btn = gr.Button("🔍 Analyze Video", variant="primary")
+                analysis_out = gr.Textbox(label="📊 Analysis Report", lines=30, show_copy_button=True)
+                analyze_btn.click(
+                    analyze_with_cookies, inputs=[url_in, cookies_in], outputs=analysis_out, show_progress=True
+                )
+            # 다운로드 탭
             with gr.TabItem("⬇️ Video Download"):
                 with gr.Row():
                     quality_dd = gr.Dropdown(
-                        choices=["best", "720p", "480p"],
-                        value="best",
-                        label="📺 Quality",
+                        choices=["best", "720p", "480p"], value="best", label="📺 Quality"
                     )
                     audio_cb = gr.Checkbox(label="🎵 Audio only (MP3)")
                 download_btn = gr.Button("⬇️ Download Video", variant="primary")
-                dl_status = gr.Textbox(
-                    label="📥 Download Status", lines=5, show_copy_button=True
-                )
+                dl_status = gr.Textbox(label="📥 Download Status", lines=5, show_copy_button=True)
                 dl_file = gr.File(label="📁 Downloaded File", visible=False)
 
-                def wrapped_download(url, q, a, cfile, progress=gr.Progress()):
-                    fp, st = download_with_cookies(url, q, a, cfile, progress)
-                    if fp and os.path.exists(fp):
-                        return st, gr.update(value=fp, visible=True)
-                    return st, gr.update(visible=False)
+                def wrapped_dl(u, q, a, c, prog=gr.Progress()):
+                    fp, st = download_with_cookies(u, q, a, c, prog)
+                    return (st, gr.update(value=fp, visible=True)) if fp and os.path.exists(fp) else (
+                        st,
+                        gr.update(visible=False),
+                    )
 
                 download_btn.click(
-                    fn=wrapped_download,
+                    wrapped_dl,
                     inputs=[url_in, quality_dd, audio_cb, cookies_in],
                     outputs=[dl_status, dl_file],
                     show_progress=True,
                 )
+            # 자막 탭  NEW
+            with gr.TabItem("🗒️ Transcript"):
+                tr_btn = gr.Button("📜 Get Full Transcript", variant="primary")
+                tr_out = gr.Textbox(
+                    label="🗒️ Transcript (full)", lines=30, show_copy_button=True
+                )
+                tr_btn.click(
+                    get_transcript, inputs=[url_in, cookies_in], outputs=tr_out, show_progress=True
+                )
 
-        # API 버튼 동작
-        api_btn.click(
-            fn=configure_api_key,
-            inputs=[api_key_in],
-            outputs=[api_status],
-        )
+        # API 버튼
+        api_btn.click(configure_api_key, inputs=[api_key_in], outputs=[api_status])
 
         gr.HTML(
             """
 <div style="margin-top:20px;padding:15px;background:#f0f8ff;border-left:5px solid #4285f4;border-radius:10px;">
   <h3>💡 Tip: 쿠키 파일 자동 사용</h3>
   <p><code>www.youtube.com_cookies.txt</code> 파일을 <strong>app.py</strong>와 같은
-  폴더에 두면 자동으로 사용됩니다. 주기적으로 새 파일로 교체해 주세요.</p>
+  폴더에 두면 업로드 없이 자동 사용됩니다.</p>
 </div>
 """
         )
@@ -718,7 +1607,7 @@ def create_interface():
 
 
 # =================================================================
-# Entrypoint
+# 실행
 # =================================================================
 if __name__ == "__main__":
     demo = create_interface()