diff --git a/video-ai-system/lib/cli.py b/video-ai-system/lib/cli.py
new file mode 100644
index 0000000..fc9f081
--- /dev/null
+++ b/video-ai-system/lib/cli.py
@@ -0,0 +1,248 @@
+# LIB · 蛋蛋短剧生产管线 · 命令行入口
+# D180 · 2026-07-10
+# Usage: python -m lib.cli <command> [args]
+import sys
+import os
+sys.path.insert(0, os.path.dirname(os.path.dirname(__file__)))
+
+from lib.models import get_db, init
+
+def cmd_init():
+    init()
+
+def cmd_shot_list(episode="DSV-EP01"):
+    db = get_db()
+    rows = db.execute("""
+        SELECT shot_number, description, status, model, seedance_task_id, output_path, error_log
+        FROM shots WHERE episode_id = ? ORDER BY shot_number
+    """, (episode,)).fetchall()
+    if not rows:
+        print(f"No shots found for {episode}. Run: python -m lib.cli sync-shots")
+        return
+    print(f"\n{'#':4s} {'状态':10s} {'描述':40s} {'模型':25s} {'任务ID':30s}")
+    print("-" * 115)
+    for r in rows:
+        status_icon = {"pending":"🔴","generating":"🟡","done":"✅","failed":"❌"}.get(r["status"], "❓")
+        desc = (r["description"] or "")[:38]
+        task = (r["seedance_task_id"] or "")[:28]
+        print(f"S{r['shot_number']:>2s}  {status_icon} {r['status']:8s}  {desc:40s} {r['model']:25s} {task:30s}")
+
+def cmd_shot_status(shot_id):
+    db = get_db()
+    r = db.execute("""
+        SELECT * FROM shots WHERE id = ?
+    """, (shot_id,)).fetchone()
+    if not r:
+        print(f"Shot {shot_id} not found")
+        return
+    print(f"\nShot {r['shot_number']} · {r['status']}")
+    print(f"  描述: {r['description']}")
+    print(f"  剧本: {r['script_ref']}")
+    print(f"  镜头: {r['camera']} · {r['duration_sec']}s")
+    print(f"  模型: {r['model']} · {r['resolution']}")
+    print(f"  提示词: {r['prompt']}")
+    print(f"  任务ID: {r['seedance_task_id']}")
+    print(f"  输出: {r['output_path']}")
+    if r['error_log']:
+        print(f"  ❌ 错误: {r['error_log']}")
+    # Show linked assets
+    assets = db.execute("""
+        SELECT a.type, a.name, sa.role FROM shot_assets sa
+        JOIN assets a ON sa.asset_id = a.id WHERE sa.shot_id = ?
+    """, (shot_id,)).fetchall()
+    if assets:
+        print(f"  关联资产:")
+        for a in assets:
+            print(f"    [{a['role']}] {a['type']}: {a['name']}")
+
+def cmd_asset_list(project="deep-sea-voyage"):
+    db = get_db()
+    rows = db.execute("""
+        SELECT id, type, name, status, approved_at FROM assets
+        WHERE project_id = ? ORDER BY type, name
+    """, (project,)).fetchall()
+    print(f"\n{'类型':6s} {'名称':25s} {'状态':10s} {'批准时间':20s} {'ID'}")
+    print("-" * 75)
+    for r in rows:
+        s = "✅" if r["status"] == "approved" else "🔴"
+        print(f"{r['type']:6s} {r['name']:25s} {s} {r['status']:8s} {(r['approved_at'] or ''):20s} {r['id']}")
+
+def cmd_sync_shots():
+    """Import shots from SHOT-LIST-EP01.hdlp into database"""
+    db = get_db()
+    shot_list_path = os.path.join(
+        os.path.dirname(os.path.dirname(__file__)),
+        "projects", "deep-sea-voyage", "shots", "SHOT-LIST-EP01.hdlp"
+    )
+    if not os.path.exists(shot_list_path):
+        print(f"Shot list not found: {shot_list_path}")
+        return
+
+    with open(shot_list_path, 'r', encoding='utf-8') as f:
+        content = f.read()
+
+    # Parse shots - simple regex approach
+    import re
+    shots = []
+    blocks = re.split(r'### Shot (\d+)', content)
+    for i in range(1, len(blocks), 2):
+        num = blocks[i]
+        text = blocks[i+1] if i+1 < len(blocks) else ""
+        shots.append((num, text.strip()))
+
+    for num, text in shots:
+        shot_id = f"DSV-EP01-S{num.zfill(2)}"
+        # Extract description
+        desc_match = re.search(r'\*\*剧本\*\*[：:]\s*(.+?)(?:\n|$)', text)
+        desc = desc_match.group(1)[:80] if desc_match else f"Shot {num}"
+        # Extract duration
+        dur_match = re.search(r'(\d+)s', text.split('\n')[0] if '\n' in text else text)
+        dur = float(dur_match.group(1)) if dur_match else 6.0
+        # Determine status
+        status = "done" if num == "01" else "pending"
+
+        db.execute("""
+            INSERT OR REPLACE INTO shots
+            (id, episode_id, shot_number, description, script_ref, camera, duration_sec, status, output_path)
+            VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
+        """, (
+            shot_id, "DSV-EP01", f"S{num.zfill(2)}",
+            desc, f"EP01-SCRIPT-LOCK line {num}",
+            "9:16 vertical", dur, status,
+            f"outputs/videos/SHOT-{num.zfill(2)}.mp4" if status == "done" else None
+        ))
+    db.commit()
+    print(f"Synced {len(shots)} shots from SHOT-LIST-EP01.hdlp")
+
+def cmd_link_asset(shot_id, asset_id, role):
+    db = get_db()
+    db.execute("INSERT OR REPLACE INTO shot_assets (shot_id, asset_id, role) VALUES (?, ?, ?)",
+               (shot_id, asset_id, role))
+    db.commit()
+    print(f"Linked: {shot_id} ← [{role}] {asset_id}")
+
+def cmd_task_log(shot_id=None):
+    db = get_db()
+    if shot_id:
+        rows = db.execute("""
+            SELECT * FROM generation_tasks WHERE shot_id = ? ORDER BY created_at DESC
+        """, (shot_id,)).fetchall()
+    else:
+        rows = db.execute("""
+            SELECT * FROM generation_tasks ORDER BY created_at DESC LIMIT 20
+        """).fetchall()
+    for r in rows:
+        icon = {"submitted":"📤","running":"🔄","succeeded":"✅","failed":"❌"}.get(r["status"], "❓")
+        print(f"{icon} {r['id'][:20]:20s} {r['shot_id']:15s} {r['task_type']:12s} {r['status']}")
+
+def cmd_exp_add(category, title, content):
+    db = get_db()
+    db.execute("INSERT INTO experience_log (category, title, content) VALUES (?, ?, ?)",
+               (category, title, content))
+    db.commit()
+    print(f"Experience logged: [{category}] {title}")
+
+def cmd_exp_list(category=None):
+    db = get_db()
+    if category:
+        rows = db.execute("SELECT * FROM experience_log WHERE category = ? ORDER BY created_at DESC",
+                         (category,)).fetchall()
+    else:
+        rows = db.execute("SELECT * FROM experience_log ORDER BY created_at DESC LIMIT 20").fetchall()
+    for r in rows:
+        print(f"[{r['category']}] {r['title']} ({r['created_at']})")
+        print(f"  {r['content'][:120]}")
+
+def cmd_chat(prompt):
+    from lib.doubao_chat import chat
+    r = chat(prompt)
+    if "content" in r:
+        print(r["content"])
+    else:
+        print(f"❌ {r}")
+
+def cmd_breakdown(file_path, episode_num=1):
+    from lib.doubao_chat import breakdown_script
+    with open(file_path, "r", encoding="utf-8") as f:
+        script = f.read()
+    r = breakdown_script(script, int(episode_num))
+    if "content" in r:
+        print(r["content"])
+        print(f"\n📊 {r['model']} · {r['tokens'].get('total_tokens',0)} tokens")
+    else:
+        print(f"❌ {r}")
+
+COMMANDS = {
+    "init": cmd_init,
+    "shots": cmd_shot_list,
+    "shot": cmd_shot_status,
+    "assets": cmd_asset_list,
+    "sync-shots": cmd_sync_shots,
+    "link": cmd_link_asset,
+    "tasks": cmd_task_log,
+    "exp-add": cmd_exp_add,
+    "exp-list": cmd_exp_list,
+    "chat": cmd_chat,
+    "breakdown": cmd_breakdown,
+    "gen": cmd_generate,
+    "collect": cmd_collect,
+}
+
+def cmd_generate(shot_id):
+    """提交 Seedance 生成任务（异步）"""
+    from lib.seedance import generate_shot_async
+    from lib.models import get_db
+    db = get_db()
+    shot = db.execute("SELECT * FROM shots WHERE id = ?", (shot_id,)).fetchone()
+    if not shot:
+        print(f"镜头 {shot_id} 不存在")
+        return
+    if shot["status"] == "done":
+        print(f"⚠️ {shot_id} 已完成，需爸爸确认是否重跑")
+        return
+
+    # 获取关联资产
+    refs = db.execute("""
+        SELECT sa.role, a.file_path FROM shot_assets sa
+        JOIN assets a ON sa.asset_id = a.id WHERE sa.shot_id = ?
+    """, (shot_id,)).fetchall()
+    ref_paths = [(r["role"], r["file_path"]) for r in refs]
+
+    prompt = shot["prompt"] or shot["description"]
+    if not prompt:
+        print(f"❌ {shot_id} 没有提示词，请先设置")
+        return
+
+    r = generate_shot_async(shot_id, prompt, ref_paths, int(shot["duration_sec"] or 6))
+    if "error" in r:
+        print(f"❌ {r['error']}")
+    else:
+        print(f"📤 {shot_id} 已提交 · task_id: {r['task_id'][:20]}...")
+        print(f"   轮询: python -m lib.cli collect {shot_id}")
+
+def cmd_collect(shot_id):
+    """收集中间结果（检查+下载）"""
+    from lib.seedance import check_and_collect
+    r = check_and_collect(shot_id)
+    if r.get("error"):
+        print(f"❌ {r['error']}")
+    elif r.get("status") == "done":
+        print(f"✅ {shot_id} 完成 → {r['output_path']}")
+    elif r.get("status") == "running":
+        print(f"🔄 {shot_id} 生成中（ID: {r['task_id'][:20]}...）")
+    else:
+        print(f"❓ {r}")
+
+if __name__ == "__main__":
+    if len(sys.argv) < 2 or sys.argv[1] not in COMMANDS:
+        print("LIB · 蛋蛋短剧生产管线 · D180")
+        print("Usage: python -m lib.cli <command> [args]")
+        print(f"Commands: {', '.join(COMMANDS.keys())}")
+        sys.exit(1)
+    cmd = COMMANDS[sys.argv[1]]
+    args = sys.argv[2:]
+    try:
+        cmd(*args)
+    except TypeError as e:
+        print(f"参数错误: {e}")
+        print(f"Usage: python -m lib.cli {sys.argv[1]} <args>")
diff --git a/video-ai-system/lib/doubao_chat.py b/video-ai-system/lib/doubao_chat.py
new file mode 100644
index 0000000..1cb665d
--- /dev/null
+++ b/video-ai-system/lib/doubao_chat.py
@@ -0,0 +1,123 @@
+# LIB · 豆包对话模型适配器（剧本拆解/分镜生成）
+# D180 · 2026-07-10
+"""火山方舟 ARK Chat API → doubao-seed 系列模型"""
+import json
+import os
+import subprocess
+import tempfile
+
+ARK_KEY = "ark-ddeba9f4-8c5a-449e-b549-9c29ec1e6f8c-a39ea"
+ARK_CHAT_URL = "https://ark.cn-beijing.volces.com/api/v3/chat/completions"
+
+# 可用模型
+MODELS = {
+    "pro": "doubao-seed-2-1-pro-260628",     # 最强·适合复杂剧本
+    "lite": "doubao-seed-2-0-lite-260215",    # 轻量·适合批量
+    "deepseek": "deepseek-v4-pro-260425",     # 深度思���·适合拆解
+}
+
+def _curl_api(payload):
+    """通过 subprocess curl 调用 API（绕过 Windows urllib 超时问题）"""
+    import subprocess, tempfile
+    tf = tempfile.NamedTemporaryFile(mode='w', suffix='.json', delete=False, encoding='utf-8')
+    json.dump(payload, tf, ensure_ascii=False)
+    tf.close()
+    try:
+        result = subprocess.run([
+            "curl", "-s", ARK_CHAT_URL,
+            "-H", f"Authorization: Bearer {ARK_KEY}",
+            "-H", "Content-Type: application/json",
+            "-d", f"@{tf.name}"
+        ], capture_output=True, text=True, timeout=120)
+        os.unlink(tf.name)
+        return json.loads(result.stdout) if result.stdout else {"error": result.stderr}
+    except subprocess.TimeoutExpired:
+        os.unlink(tf.name)
+        return {"error": "timeout"}
+
+def chat(prompt, system="你是一个专业的短剧剧本分析专家", model="pro", temperature=0.3, max_tokens=4096):
+    """调用豆包对话模型"""
+    payload = {
+        "model": MODELS.get(model, model),
+        "messages": [
+            {"role": "system", "content": system},
+            {"role": "user", "content": prompt}
+        ],
+        "temperature": temperature,
+        "max_tokens": max_tokens
+    }
+    result = _curl_api(payload)
+    if "error" in result:
+        return result
+    choice = result.get("choices", [{}])[0]
+    return {
+        "content": choice.get("message", {}).get("content", ""),
+        "model": result.get("model", ""),
+        "tokens": result.get("usage", {}),
+        "finish": choice.get("finish_reason", "")
+    }
+
+def breakdown_script(script_text, episode_num=1):
+    """拆解一集剧本为结构化分镜"""
+    prompt = f"""请将以下短剧剧本第{episode_num}集拆解为结构化分镜。 
+
+要求：
+1. 按镜头拆分，每个镜头包含：镜头编号、景别（特写/近景/中景/全景/POV）、时长（秒）
+2. 提取每个镜头中出现的角色、场景、道具
+3. 写出每个镜头的画面描述（50字以内）
+4. 标注镜头类型：establishing/action/reaction/closeup/transition
+
+输出JSON格式：
+{{
+  "episode": {episode_num},
+  "shots": [
+    {{
+      "shot_number": "S01",
+      "description": "画面描述",
+      "camera": "POV|特写|近景|中景|全景",
+      "duration": 6,
+      "type": "establishing|action|reaction|closeup|transition",
+      "characters": ["角色名"],
+      "scenes": ["场景名"],
+      "props": ["道具名"],
+      "dialogue": null
+    }}
+  ]
+}}
+
+剧本内容：
+{script_text}"""
+    return chat(prompt, system="你是专业的短剧剧本拆解专家，输出纯JSON，不添加任何解释。")
+
+def generate_keyframes(shot_description, character_refs, scene_refs):
+    """为单个镜头生成关键帧 prompt"""
+    prompt = f"""基于以下镜头描述，生成即梦4.0图像生成提示词。
+    
+镜头：{shot_description}
+可用角色：{json.dumps(character_refs, ensure_ascii=False)}
+可用场景：{json.dumps(scene_refs, ensure_ascii=False)}
+
+要求：
+1. 提示词用英文
+2. 包含景别、角色位置、场景细节、光照、风格
+3. 不超过150词
+4. 输出纯提示词，不加任何解释"""
+    return chat(prompt, model="pro", temperature=0.5, max_tokens=300)
+
+# ====== CLI ======
+if __name__ == "__main__":
+    import sys
+    if len(sys.argv) < 2:
+        print("Usage: python doubao_chat.py <command> [args]")
+        print("  chat <prompt>    直接对话")
+        print("  breakdown <file> 拆解剧本文件")
+        sys.exit(1)
+    cmd = sys.argv[1]
+    if cmd == "chat":
+        r = chat(sys.argv[2])
+        print(r["content"] if "content" in r else r)
+    elif cmd == "breakdown":
+        with open(sys.argv[2], "r", encoding="utf-8") as f:
+            script = f.read()
+        r = breakdown_script(script)
+        print(r["content"] if "content" in r else r)
diff --git a/video-ai-system/lib/models.py b/video-ai-system/lib/models.py
new file mode 100644
index 0000000..e3d183f
--- /dev/null
+++ b/video-ai-system/lib/models.py
@@ -0,0 +1,109 @@
+# LIB · 蛋蛋短剧生产管线 · 数据库模型
+# D180 · 2026-07-10
+import sqlite3
+import json
+import os
+from datetime import datetime, timezone
+
+DB_PATH = os.path.join(os.path.dirname(__file__), "lib.db")
+
+SCHEMA = """
+CREATE TABLE IF NOT EXISTS projects (
+    id TEXT PRIMARY KEY,
+    name TEXT NOT NULL,
+    created_at TEXT DEFAULT (datetime('now')),
+    updated_at TEXT DEFAULT (datetime('now'))
+);
+
+CREATE TABLE IF NOT EXISTS episodes (
+    id TEXT PRIMARY KEY,
+    project_id TEXT NOT NULL REFERENCES projects(id),
+    number INTEGER NOT NULL,
+    script_path TEXT,
+    status TEXT DEFAULT 'pending',
+    created_at TEXT DEFAULT (datetime('now'))
+);
+
+CREATE TABLE IF NOT EXISTS shots (
+    id TEXT PRIMARY KEY,
+    episode_id TEXT NOT NULL REFERENCES episodes(id),
+    shot_number TEXT NOT NULL,
+    description TEXT,
+    script_ref TEXT,
+    camera TEXT,
+    duration_sec REAL,
+    status TEXT DEFAULT 'pending',
+    prompt TEXT,
+    model TEXT DEFAULT 'doubao-seedance-2-0-260128',
+    resolution TEXT DEFAULT '720p',
+    output_path TEXT,
+    seedance_task_id TEXT,
+    created_at TEXT DEFAULT (datetime('now')),
+    updated_at TEXT DEFAULT (datetime('now')),
+    error_log TEXT
+);
+
+CREATE TABLE IF NOT EXISTS assets (
+    id TEXT PRIMARY KEY,
+    project_id TEXT NOT NULL REFERENCES projects(id),
+    type TEXT NOT NULL CHECK(type IN ('CHAR','ENV','PROP','MASTER')),
+    name TEXT NOT NULL,
+    file_path TEXT NOT NULL,
+    status TEXT DEFAULT 'pending',
+    approved_at TEXT,
+    created_at TEXT DEFAULT (datetime('now'))
+);
+
+CREATE TABLE IF NOT EXISTS shot_assets (
+    shot_id TEXT NOT NULL REFERENCES shots(id),
+    asset_id TEXT NOT NULL REFERENCES assets(id),
+    role TEXT NOT NULL CHECK(role IN ('reference','character','scene','prop','master')),
+    PRIMARY KEY (shot_id, asset_id, role)
+);
+
+CREATE TABLE IF NOT EXISTS generation_tasks (
+    id TEXT PRIMARY KEY,
+    shot_id TEXT NOT NULL REFERENCES shots(id),
+    task_type TEXT NOT NULL CHECK(task_type IN ('image','video','master_frame')),
+    seedance_task_id TEXT,
+    status TEXT DEFAULT 'submitted',
+    prompt_used TEXT,
+    references_used TEXT,
+    error_message TEXT,
+    started_at TEXT,
+    completed_at TEXT,
+    created_at TEXT DEFAULT (datetime('now'))
+);
+
+CREATE TABLE IF NOT EXISTS experience_log (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    category TEXT NOT NULL,
+    title TEXT NOT NULL,
+    content TEXT NOT NULL,
+    tags TEXT,
+    created_at TEXT DEFAULT (datetime('now'))
+);
+"""
+
+def get_db():
+    db = sqlite3.connect(DB_PATH)
+    db.row_factory = sqlite3.Row
+    db.executescript(SCHEMA)
+    db.commit()
+    return db
+
+def init():
+    db = get_db()
+    # Create default project and episode if not exist
+    db.execute("INSERT OR IGNORE INTO projects (id, name) VALUES (?, ?)",
+               ("deep-sea-voyage", "深海迷航"))
+    db.execute("""INSERT OR IGNORE INTO episodes (id, project_id, number, script_path, status)
+                  VALUES (?, ?, ?, ?, ?)""",
+               ("DSV-EP01", "deep-sea-voyage", 1,
+                "projects/deep-sea-voyage/script/EP01-SCRIPT-LOCK.hdlp", "ready"))
+    db.commit()
+    print(f"LIB initialized: {DB_PATH}")
+    return db
+
+if __name__ == "__main__":
+    init()
diff --git a/video-ai-system/lib/seedance.py b/video-ai-system/lib/seedance.py
new file mode 100644
index 0000000..754b22b
--- /dev/null
+++ b/video-ai-system/lib/seedance.py
@@ -0,0 +1,149 @@
+# LIB · Seedance 2.0 视频生成适配器 + 异步任务队列
+# D180 · 2026-07-10
+"""火山方舟 ARK → Seedance 2.0 · 提交+轮询+下载+写库 全自动"""
+import json, os, subprocess, time, tempfile, base64
+
+ARK_KEY = "ark-ddeba9f4-8c5a-449e-b549-9c29ec1e6f8c-a39ea"
+ARK_BASE = "https://ark.cn-beijing.volces.com/api/v3"
+SEEDANCE_MODEL = "doubao-seedance-2-0-260128"
+TASK_URL = f"{ARK_BASE}/contents/generations/tasks"
+
+def _curl(method, url, payload=None):
+    """通用 curl 调用"""
+    tf = None
+    cmd = ["curl", "-s", url, "-H", f"Authorization: Bearer {ARK_KEY}"]
+    if payload:
+        tf = tempfile.NamedTemporaryFile(mode='w', suffix='.json', delete=False, encoding='utf-8')
+        json.dump(payload, tf, ensure_ascii=False)
+        tf.close()
+        cmd += ["-H", "Content-Type: application/json", "-d", f"@{tf.name}"]
+    if method == "POST":
+        cmd.insert(1, "-X")
+        cmd.insert(2, "POST")
+    try:
+        r = subprocess.run(cmd, capture_output=True, text=True, timeout=15)
+        if tf: os.unlink(tf.name)
+        return json.loads(r.stdout) if r.stdout else {"error": r.stderr}
+    except Exception as e:
+        if tf: os.unlink(tf.name)
+        return {"error": str(e)}
+
+def submit_task(prompt, references=None, duration=6, resolution="720p"):
+    """提交 Seedance 生成任务 → 返回 task_id"""
+    content = [{"type": "text", "text": prompt}]
+    if references:
+        for ref_type, ref_path in references:
+            with open(ref_path, "rb") as f:
+                fmt = "png" if ref_path.endswith(".png") else "jpeg"
+                b64 = base64.b64encode(f.read()).decode()
+            content.append({
+                "type": "image_url",
+                "role": ref_type,
+                "image_url": {"url": f"data:image/{fmt};base64,{b64}"}
+            })
+    payload = {
+        "model": SEEDANCE_MODEL,
+        "content": content,
+        "duration": duration,
+        "resolution": resolution
+    }
+    result = _curl("POST", TASK_URL, payload)
+    if "id" in result:
+        return result["id"]
+    return {"error": result}
+
+def poll_task(task_id, poll_interval=30, max_wait=600):
+    """轮询任务状态 → 返回结果"""
+    elapsed = 0
+    while elapsed < max_wait:
+        result = _curl("GET", f"{TASK_URL}/{task_id}")
+        if "error" in result:
+            return result
+        status = result.get("status", "")
+        if status == "succeeded":
+            return result
+        if status == "failed":
+            return {"error": f"Task failed: {result}"}
+        time.sleep(poll_interval)
+        elapsed += poll_interval
+    return {"error": "timeout"}
+
+def download_video(video_url, output_path):
+    """下载视频"""
+    subprocess.run(["curl", "-s", "-o", output_path, video_url], check=False)
+    return os.path.exists(output_path) and os.path.getsize(output_path) > 0
+
+def generate_shot(shot_id, prompt, references, duration=6, output_dir="outputs/videos"):
+    """一站式生成：提交→轮询→下载→返回结果"""
+    print(f"[LIB] 提交 {shot_id}...")
+    os.makedirs(output_dir, exist_ok=True)
+    task_id = submit_task(prompt, references, duration)
+    if isinstance(task_id, dict):
+        return {"error": task_id}
+
+    print(f"[LIB] 任务 {task_id[:20]}... 等待完成")
+    result = poll_task(task_id)
+    if "error" in result:
+        return result
+
+    video_url = result.get("content", {}).get("video_url")
+    if not video_url:
+        return {"error": "no video_url"}
+
+    output_path = os.path.join(output_dir, f"{shot_id}.mp4")
+    print(f"[LIB] 下载视频...")
+    if download_video(video_url, output_path):
+        return {
+            "task_id": task_id,
+            "output_path": output_path,
+            "seed": result.get("seed"),
+            "duration": result.get("duration"),
+            "resolution": result.get("resolution"),
+            "tokens": result.get("usage", {}).get("total_tokens", 0)
+        }
+    return {"error": "download failed"}
+
+def generate_shot_async(shot_id, prompt, references, duration=6, output_dir="outputs/videos"):
+    """异步提交：只提交不等待，返回 task_id 用于后续轮询"""
+    task_id = submit_task(prompt, references, duration)
+    if isinstance(task_id, dict):
+        return task_id
+    # 写入任务记录
+    from lib.models import get_db
+    db = get_db()
+    db.execute("""
+        INSERT INTO generation_tasks (id, shot_id, task_type, seedance_task_id, status, prompt_used, references_used, started_at)
+        VALUES (?, ?, ?, ?, ?, ?, ?, datetime('now'))
+    """, (task_id[:20], shot_id, "video", task_id, "running", prompt, json.dumps([r[0] for r in references] if references else [])))
+    db.execute("UPDATE shots SET seedance_task_id = ?, status = 'generating' WHERE id = ?", (task_id, shot_id))
+    db.commit()
+    return {"task_id": task_id, "status": "submitted"}
+
+def check_and_collect(shot_id):
+    """检查任务状态，完成后自动下载"""
+    from lib.models import get_db
+    db = get_db()
+    task = db.execute("""
+        SELECT * FROM generation_tasks WHERE shot_id = ? ORDER BY created_at DESC LIMIT 1
+    """, (shot_id,)).fetchone()
+    if not task:
+        return {"error": "no task found"}
+    result = poll_task(task["seedance_task_id"], poll_interval=0, max_wait=1)
+    if "error" in result:
+        return result
+    if "video_url" not in str(result):
+        return {"status": "running", "task_id": task["seedance_task_id"]}
+
+    # Succeeded - download
+    video_url = result.get("content", {}).get("video_url")
+    output_dir = os.path.join(os.path.dirname(os.path.dirname(__file__)), "outputs", "videos")
+    os.makedirs(output_dir, exist_ok=True)
+    output_path = os.path.join(output_dir, f"{shot_id}.mp4")
+    if download_video(video_url, output_path):
+        db.execute("""
+            UPDATE generation_tasks SET status = 'succeeded', completed_at = datetime('now') WHERE id = ?
+        """, (task["id"],))
+        db.execute("UPDATE shots SET status = 'done', output_path = ? WHERE id = ?", (output_path, shot_id))
+        db.commit()
+        return {"status": "done", "output_path": output_path, "seed": result.get("seed")}
+    return {"error": "download failed"}
diff --git a/video-ai-system/lib/seedream.py b/video-ai-system/lib/seedream.py
new file mode 100644
index 0000000..491907b
--- /dev/null
+++ b/video-ai-system/lib/seedream.py
@@ -0,0 +1,104 @@
+# LIB · 即梦4.0 (Seedream 4.0) 适配器
+# D180 · 2026-07-10
+"""火山方舟 ARK API → Seedream 4.0 图像生成"""
+import json
+import os
+import urllib.request
+import base64
+
+ARK_KEY = "ark-ddeba9f4-8c5a-449e-b549-9c29ec1e6f8c-a39ea"
+ARK_BASE = "https://ark.cn-beijing.volces.com/api/v3"
+SEEDREAM_MODEL = "doubao-seedream-4-0-250828"
+
+def _api_call(endpoint, payload):
+    """通用 ARK API 调用"""
+    req = urllib.request.Request(
+        f"{ARK_BASE}{endpoint}",
+        data=json.dumps(payload).encode(),
+        headers={
+            "Authorization": f"Bearer {ARK_KEY}",
+            "Content-Type": "application/json"
+        },
+        method="POST"
+    )
+    try:
+        with urllib.request.urlopen(req, timeout=30) as resp:
+            return json.loads(resp.read())
+    except urllib.error.HTTPError as e:
+        err_body = e.read().decode()
+        return {"error": str(e.code), "detail": err_body}
+
+def seedream_t2i(prompt, size="1440x2560", n=1):
+    """文生图"""
+    payload = {
+        "model": SEEDREAM_MODEL,
+        "prompt": prompt,
+        "size": size,
+        "n": n
+    }
+    return _api_call("/images/generations", payload)
+
+def seedream_i2i(prompt, image_path, size="1440x2560", n=1):
+    """图生图"""
+    with open(image_path, "rb") as f:
+        img_b64 = base64.b64encode(f.read()).decode()
+    payload = {
+        "model": SEEDREAM_MODEL,
+        "prompt": prompt,
+        "image": f"data:image/jpeg;base64,{img_b64}",
+        "size": size,
+        "n": n
+    }
+    return _api_call("/images/generations", payload)
+
+def seedream_multi(prompt, image_paths, size="1440x2560", n=1):
+    """多图融合"""
+    imgs = []
+    for p in image_paths:
+        with open(p, "rb") as f:
+            fmt = "png" if p.endswith(".png") else "jpeg"
+            imgs.append(f"data:image/{fmt};base64,{base64.b64encode(f.read()).decode()}")
+    payload = {
+        "model": SEEDREAM_MODEL,
+        "prompt": prompt,
+        "image": imgs,
+        "size": size,
+        "n": n
+    }
+    return _api_call("/images/generations", payload)
+
+def seedream_download(result, output_path):
+    """从 API 结果下载图片"""
+    if "error" in result:
+        return None
+    url = result.get("data", [{}])[0].get("url")
+    if not url:
+        return None
+    urllib.request.urlretrieve(url, output_path)
+    return output_path
+
+# ====== CLI ======
+if __name__ == "__main__":
+    import sys
+    if len(sys.argv) < 2:
+        print("Usage: python seedream.py t2i <prompt> [output]")
+        print("       python seedream.py i2i <prompt> <image_path> [output]")
+        print("       python seedream.py multi <prompt> <img1> <img2> [output]")
+        sys.exit(1)
+    cmd = sys.argv[1]
+    if cmd == "t2i":
+        prompt = sys.argv[2]
+        out = sys.argv[3] if len(sys.argv) > 3 else "seedream_output.jpg"
+        r = seedream_t2i(prompt)
+        p = seedream_download(r, out)
+        print(f"✅ {p}" if p else f"❌ {r.get('error','unknown')}")
+    elif cmd == "i2i":
+        prompt, img, out = sys.argv[2], sys.argv[3], (sys.argv[4] if len(sys.argv) > 4 else "seedream_i2i_output.jpg")
+        r = seedream_i2i(prompt, img)
+        p = seedream_download(r, out)
+        print(f"✅ {p}" if p else f"❌ {r.get('error','unknown')}")
+    elif cmd == "multi":
+        prompt, imgs, out = sys.argv[2], sys.argv[3:-1], sys.argv[-1]
+        r = seedream_multi(prompt, imgs)
+        p = seedream_download(r, out)
+        print(f"✅ {p}" if p else f"❌ {r.get('error','unknown')}")
diff --git a/video-ai-system/lib/tts.py b/video-ai-system/lib/tts.py
new file mode 100644
index 0000000..637f878
--- /dev/null
+++ b/video-ai-system/lib/tts.py
@@ -0,0 +1,74 @@
+# LIB · TTS 配音模块
+# D180 · 2026-07-10
+"""Edge-TTS 配音引擎 · 多角色音色"""
+import asyncio, edge_tts, os
+
+# 角色音色表
+VOICES = {
+    "林昊": "zh-CN-YunxiNeural",      # 青年男声·迷茫虚弱
+    "旁白": "zh-CN-YunyangNeural",     # 新闻男声
+    "默认女": "zh-CN-XiaoxiaoNeural",  # 女声
+    "默认男": "zh-CN-YunxiNeural",
+}
+
+FFMPEG = os.path.expanduser("~/.workbuddy/binaries/ffmpeg.exe")
+
+def tts(text, output_path, voice="zh-CN-YunxiNeural", rate="+0%"):
+    """生成语音 mp3"""
+    os.makedirs(os.path.dirname(output_path) or ".", exist_ok=True)
+    async def _run():
+        await edge_tts.Communicate(text, voice, rate=rate).save(output_path)
+    asyncio.run(_run())
+    return output_path if os.path.exists(output_path) else None
+
+def tts_for_character(text, character, output_path, rate="+0%"):
+    """按角色名查音色生成"""
+    voice = VOICES.get(character, "zh-CN-YunxiNeural")
+    return tts(text, output_path, voice, rate)
+
+def mix_audio_video(video_path, audio_path, output_path, video_vol=0.3):
+    """FFmpeg 合成视频+音频"""
+    if not os.path.exists(FFMPEG):
+        return f"ffmpeg not found at {FFMPEG}"
+    os.makedirs(os.path.dirname(output_path) or ".", exist_ok=True)
+    import subprocess
+    r = subprocess.run([
+        FFMPEG, "-y",
+        "-i", video_path,
+        "-i", audio_path,
+        "-filter_complex", f"[0:a]volume={video_vol}[va];[1:a]volume=1.0[aa];[va][aa]amix=inputs=2:duration=first",
+        "-c:v", "copy",
+        output_path
+    ], capture_output=True, text=True)
+    return output_path if r.returncode == 0 else r.stderr
+
+def add_subtitle(video_path, subtitle_text, output_path):
+    """FFmpeg 添加字幕"""
+    if not os.path.exists(FFMPEG):
+        return f"ffmpeg not found at {FFMPEG}"
+    import subprocess
+    r = subprocess.run([
+        FFMPEG, "-y",
+        "-i", video_path,
+        "-vf", f"drawtext=text='{subtitle_text}':fontsize=28:fontcolor=white:borderw=2:bordercolor=black:x=(w-text_w)/2:y=h-120",
+        "-c:a", "copy",
+        output_path
+    ], capture_output=True, text=True)
+    return output_path if r.returncode == 0 else r.stderr
+
+if __name__ == "__main__":
+    import sys
+    if len(sys.argv) < 2:
+        print("Usage: python tts.py tts <text> <output> [voice]")
+        print("       python tts.py char <text> <character> <output>")
+        print("       python tts.py mix <video> <audio> <output>")
+        print(f"\nVoices: {list(VOICES.keys())}")
+        sys.exit(1)
+    cmd = sys.argv[1]
+    if cmd == "tts":
+        voice = sys.argv[4] if len(sys.argv) > 4 else "zh-CN-YunxiNeural"
+        print(tts(sys.argv[2], sys.argv[3], voice) or "FAILED")
+    elif cmd == "char":
+        print(tts_for_character(sys.argv[2], sys.argv[3], sys.argv[4]) or "FAILED")
+    elif cmd == "mix":
+        print(mix_audio_video(sys.argv[2], sys.argv[3], sys.argv[4]) or "FAILED")