视频拍摄知识+AIGC数据预处理

视角

参考链接：https://www.polarpro.com/blogs/polarpro/filmmaking-101-types-of-camera-shots-and-angles

1. Low Angle Shot 低角度拍摄、
2. horizontal Shot 平视、
3. Dutch Angle Shot 荷兰角斜拍、
4. High Angle Shot 高角度拍摄、
5. Bird’s-eye / Aerial Shot 鸟瞰 / 航拍

运镜：

参考链接：https://www.storyblocks.com/resources/tutorials/7-basic-camera-movements

1. 变焦 Zoom （放大、缩小）
2. 平移 Pan （全景，左右旋转）
3. 倾斜 Tilt （上下旋转）
4. 小车 Dolly （相对主体前后平移）
5. 卡车 Truck （左右平移）
6. 基座 Pedestal （上下平移）
7. 焦点变换 Rack focus （画面大小不变，调整镜头以模糊一个主体，同时锐化另一个主体）
8. 固定 Fixed （镜头不动）

景别

参考链接：https://images.wikia.com/scratchpad/zh/images/f/f5/%E6%94%9D%E5%BD%B1%E9%8F%A1%E9%A0%AD%E4%B8%80%E8%A6%BD%E8%A1%A8.pdf

物体可分为：

1. 远景 Long Shot
2. 全景 Medium Long Shot、Panorama
3. 特写 Close-up
4. 微距 Macro shot

人物可细分：

1. 大远景 Extra Long Shot
2. 远景 Long Shot
3. 全景 Medium Long Shot、Panorama
4. 中景 Medium-shot
5. 近景 Medium Close-Up
6. 特写 Close-up
7. 微距 Macro shot

（就拿拍摄人物来说，远景就是把整个人和环境拍摄在画面里面，常用来展示事件发生的时间、环境、规模和气氛。往往是用来塑造一个大的场景。全景就是比远景更近一点，把人物的身体整个展示在画面里面，用来表现人物的全身动作，或者是人物之间的关系。中景就是指拍摄人物膝盖至头顶的部分，这样不仅能够展示人物的表情，而且还可以显示人物的形体动作。近景也就是拍摄人物胸部以上至头部的部位，非常有利于表现人物的面部或者是其他部门的表情、神态。包括眼角的细微动作。）

prompt test

1. 视角

prompt

there is a caption of a video,: The video begins with an aerial view of a vibrant suburban neighborhood, showcasing approximately 50 houses, 10 multi-story apartment buildings, along with distinct red-roofed and green-roofed structures amidst lush greenery and tree-lined streets. Transitioning to a historical fortification in the middle frame, the scene unveils a grand stone wall encircling a spacious green area, punctuated by scattered trees and a central prominent tree, juxtaposed against modern buildings in the backdrop. Finally, the video concludes with another aerial perspective, this time capturing a historic fortification nestled within a hilly landscape, featuring stone buildings with thatched roofs, picnic areas amidst a grassy terrain, and distant houses and trees, all illustrating the captivating blend of history and nature. Each phase offers a unique glimpse into different environments, from suburban living to historical landmarks, all captured through stunning aerial photography.

The shooting angles of a video are divided into:

1. Low Angle Shot
2. horizontal Shot
3. Dutch Angle Shot
4. High Angle Shot
5. Bird’s-eye / Aerial Shot

Which perspective does the video most belong to, please return one number only

chatgpt answer (True)

The perspective of the video belongs to option 5: Bird’s-eye / Aerial Shot.

DeepSeek-VL-7B answer (True)

The video belongs to the perspective of 5. Bird’s-eye / Aerial Shot.

True answer is Bird’s-eye / Aerial Shot

2. 运镜

prompt

there is a caption of a video,: The video begins with an aerial view of a vibrant suburban neighborhood, showcasing approximately 50 houses, 10 multi-story apartment buildings, along with distinct red-roofed and green-roofed structures amidst lush greenery and tree-lined streets. Transitioning to a historical fortification in the middle frame, the scene unveils a grand stone wall encircling a spacious green area, punctuated by scattered trees and a central prominent tree, juxtaposed against modern buildings in the backdrop. Finally, the video concludes with another aerial perspective, this time capturing a historic fortification nestled within a hilly landscape, featuring stone buildings with thatched roofs, picnic areas amidst a grassy terrain, and distant houses and trees, all illustrating the captivating blend of history and nature. Each phase offers a unique glimpse into different environments, from suburban living to historical landmarks, all captured through stunning aerial photography.

The camera movement are divided into:

1. Zoom
2. Pan
3. Tilt
4. Dolly
5. Truck
6. Pedestal
7. Rack focus
8. Fixed

Which movement does the video most belong to, please return one number only

chatgpt answer (True)

Based on the description provided, the video most likely primarily utilizes option 2: Pan.

DeepSeek-VL-7B answer (False)

The video most belongs to the movement of 8. Fixed.

True answer is Pan

3. 景别

prompt

there is a caption of a video,: The video begins with an aerial view of a vibrant suburban neighborhood, showcasing approximately 50 houses, 10 multi-story apartment buildings, along with distinct red-roofed and green-roofed structures amidst lush greenery and tree-lined streets. Transitioning to a historical fortification in the middle frame, the scene unveils a grand stone wall encircling a spacious green area, punctuated by scattered trees and a central prominent tree, juxtaposed against modern buildings in the backdrop. Finally, the video concludes with another aerial perspective, this time capturing a historic fortification nestled within a hilly landscape, featuring stone buildings with thatched roofs, picnic areas amidst a grassy terrain, and distant houses and trees, all illustrating the captivating blend of history and nature. Each phase offers a unique glimpse into different environments, from suburban living to historical landmarks, all captured through stunning aerial photography.

The scene type are divided into:

1. Extra Long Shot
2. Long Shot
3. Medium Long Shot / Panorama
4. Medium-shot
5. Medium Close-Up
6. Close-up
7. Macro shot

Which scene type does the video most belong to, please return one number only

chatgpt answer (True)

Based on the description provided, the video most likely primarily utilizes option 3: Medium Long Shot / Panorama.

DeepSeek-VL-7B answer (True)

The video most belongs to the scene type of 3. Medium Long Shot / Panorama.

True answer is Medium Long Shot / Panorama

十个视频测试结果

1. 燃烧的火焰

2. 小镇鸟瞰图

3. 多人交流

4. 情侣视频

5. 绿幕圣诞女孩

6. 三文鱼烧烤

7. 电工喝水

8. 草地人群聚会

9. 船与行人

10. 鸟立枝头