Ancient Chinese Dancer in Pink-Gold Hanfu, 15s Vertical Video

I tested recreation of x.com/PixelAigc/stat… by @PixelAigc, in Grok @imagine 

As Grok @imagine does not support using video as reference, I print screen 3 images from the original video (order is right to left, 1 is the first frame).

Then I use Grok Imagine agent mode, and pasted the prompt from the original post:
```
20岁东亚女性，女孩的形象不要变，舞姿和他们一样。粉白金渐变重工古风齐胸广袖流仙裙，裙身布满金线刺绣花卉纹样，轻薄通透的鎏金镶边纱质披帛，布料随风自然飘逸灵动，裙摆垂坠感十足，裙身点缀细碎珍珠挂饰，精致古风绣花软底布鞋，整体华丽温柔。裙子里面有内衬裤子，保证美女做抬腿动作时不会漏肉出来。舞动随风飘逸；轻盈灵动；黑色长发盘成古风发髻，搭配同色系精致头饰步摇，精致额饰，兼具明艳美感与飘逸气韵 初始姿态：单左腿站立支撑，右腿前屈，身体前倾轻靠右腿，双目低垂；右手抬至头部右上方，左手向后斜上方舒展。 1. 身体缓缓直立、微微后仰，右腿同步笔直向上抬高，双手枕于鬓角，肩部轻柔晃动，头部完整旋转一周； 2. 左手自上而下绕头一圈，右手自下而上环绕胸口一圈； 3. 身体转向左侧，视线低垂看向左下方，十指交叉挡于面部，双手缓慢向上下分开； 4. 右手大力向上甩至头顶，左手向下舒展，右腿向后划出圆弧扫动，借扫腿力道完整旋转一周； 5. 右腿落地承重，左腿后抬、小腿向内勾起，身体向后仰90度；保持姿态原地匀速旋转一圈，双眼轻闭、面带浅笑，双手手掌叠放于下巴处小幅左右轻晃； 6. 动作定格，双手同步自上而下划圆弧环绕上半身一周，左手略高于右手，视线跟随手部轨迹移动； 7. 双手拇指食指相贴、其余三指舒展，从胸前快速高举过头顶； 8. 身体转向右侧，上身微前倾，左手从头顶划弧轻落至下巴，右指尖轻扫眉梢后伸向右上头顶，手臂弯曲、手腕外压；双手轻晃腕间金饰两次，整套舞姿舒缓端庄大气。最后一个舞蹈姿势完成时，正脸缓缓回头看前方。 氛围与视听补充 画面搭配重金属鼓点配乐，节奏明快有力；体积柔光，细腻皮肤肌理，发丝、纱披帛布料飘动自然，鎏金配饰反光细腻，光影层次丰富，动态丝滑无卡顿。 负面规避词 模糊、低清、马赛克、畸形肢体、长短腿、扭曲五官、手部畸形、动作卡顿跳帧、僵硬布料、无飘动披帛、刺眼强光、杂乱多余杂物、水印文字、二次元、油画风、暗沉发黑画面、残缺饰品、光效杂乱突兀，生成15秒连贯视频。9:16。9:16。
```

Grok @imagine did a pretty good job in generating a 10 seconds video and 6 seconds video and then combined into a final 15 seconds video. Looking the 10 and 6 seconds video by itself, it is fine. However, the 15 seconds video has a clear transition problem and the background music are different!

I wish Grok @imagine to improve:
- Overall plan for final video length and decide if background music is at each clip (10s) level or should be a single music merged at the end, to ensure the consistency
- improve the transition between adjacent clips, and have a post generation check 

Anyway, here is the result from Grok @imagine.