批量制作UGC视频的AI工作流

seedance 2.5 + claude code + higgsfield CLI = UGC机器... 以下是我批量制作此类视频的工作流程： 1. 让你的代理访问一个短视频内容研究代理API，我用的是一个私有的，即将发布，但你可以使用市场上任何一个。 2. 让研究代理拉取100个推广类似你产品的视频，并制作一个链接电子表格。 3. 浏览这些链接，找到一个你认为适合你销售的产品的视频。 4. 给它访问视频分析工具的权限，higgsfield CLI内置了一个你可以使用的（另外，在接下来的33天内，你可以在higgsfield上无限使用Seedance 2.5，零积分成本）。 5. 让你的代理将视频的所有上下文提取到一个临时md文件中，包括场景、动作、人物外貌、举止、口音、语调，所有内容都以结构化格式呈现。 6. 让代理使用higgsfield CLI，利用这些上下文创建许多符合视频氛围的起始帧，选择你认为最合适的一个。 7. 让你的代理更改视频上下文，以适应你的产品，并推广你的产品，而不是原视频推广的产品。 8. 现在将这个视频上下文发送给higgsfield CLI，让它使用Seedance 2.5，并使用你的起始帧来生成它。 80%的情况下，它第一次尝试就会绝对成功。 [回复]

seedance 2.5 + claude code + higgsfield CLI = UGC machine... here's my workflow to create videos like this en masse: 1. give your agent access to a shortform content research agent API, i use a private one that i'm dropping soon, but you can use any of the ones on the market 2. ask the research agent to pull 100 videos that promote a product like yours & make a spreadsheet of links 3. go through these links & find a video that you think fits the product you sells 4. give it access to a video analysis tool, the higgsfield CLI has one built-in that you can use (also you can have UNLIMITED access to Seedance 2.5 w/ 0 credit cost on higgsfield for the next 33 days) 5. ask your agent to extract ALL context about the video into a temporary md file, including scenes, motions, appearance of the people, mannerisms, accents, tone of voice, EVERYTHING in a structured format 6. make the agent use the higgsfield CLI to use this context to create many starting frames that fit the vibe of the video, pick the one that you think fits best 7. ask your agent to change the video context to fit YOUR product & promote it instead of the product the original video was promoting 8. now send this video context to the higgsfield CLI, ask it to use Seedance 2.5 & to use your starting frame to generate it 80% of the time, it will absolutely crush it first try