Anniversary video workflow refinement

From historical assets to a video

Anniversary videos bring together historical assets and brand narratives. I refined prompts and output controls in an existing workflow that turns those assets into video segments.

Choose camera movement from image content

I wrote conditional prompts that identify the subject before selecting camera movement.

Still-life and close-up scenes favor a fixed camera; wide landscapes or moving subjects can use dynamic movement. These rules address unsuitable motion and subject drift.

I classified close-ups of credentials, logos, products, and screens as fixed-camera scenes. The generation configuration enables camerafixed, while the prompt specifies a static camera, fixed perspective, and reduced deformation.

The excerpt shows two layers of control: select the configuration from the shot type, then constrain movement through the prompt. The aim is to reduce subject drift and deformation; generated outputs still require review.

Decision rulesView excerpt

#### 2.3 特写镜头防抖设置（解决物品乱动问题）

判断逻辑：检查分镜脚本，如果镜头类型为以下任一种，必须开启固定镜头模式。

需要固定镜头的场景：
- 证件特写（工牌、名片、证书）
- Logo特写（logo、公司标识）
- 产品特写（静物、展品）
- 静止人物特写（证件照式构图、肖像特写）
- 屏幕内容特写（电脑界面、数据展示）
- 图表/文档特写

Fixed-camera configurationView excerpt

// 如果是上述场景之一
{
  "camerafixed": true, // 必须设为 true
  "prompt": "原始提示词 + , static camera, minimal motion, no deformation, fixed perspective, 静态相机, 主体静止, 只有光影变化, 固定视角, 无形变"
}

Configuration comparisonView excerpt

// 未启用固定镜头的配置
{
  "camerafixed": false,
  "prompt": "logo特写, 金属质感"
}

// 启用固定镜头并补充运动约束的配置
{
  "camerafixed": true,
  "prompt": "logo特写, 金属质感, static camera, minimal motion, no deformation, 静态相机, 主体静止, 只有光影变化"
}

Prompt and parameter content is preserved, with formatting and comments edited for clarity. The comparison illustrates configurations, not measured output results.

Keep one voice throughout the task

Selecting a voice independently for each segment can lead to inconsistent narration.

I separated voice selection from the segment loop, assigning one TTS voice identifier for the task and reusing it across subsequent segments.

Delivery and scope

My work covered camera-selection rules and narration consistency in an existing workflow. Backend developers handled website deployment after the team completed the outputs.

Back to portfolio