reddit · SideProject · 09-15 11:50 55 分 ← 返回列表

I built an open-source tool that can replicate viral video ads in minutes, not in days

A lot of people say AI video tools can produce finished videos for us. But when you actually use them, you realize there’s still a lot of manual work involved in fixing the details. Our old workflow : find a video that was performing well, use it as a reference, remark it for our product, then publish it. A single video will take an inter a whole day. We also tried to use AI agents to handle the whole process, but the results weren’t good. Tools like Seedance 2.0 and Kling m

五维评分

付费意愿

需求真实度

个人可落地

竞争空间

新鲜度

AI 分析

痛点：用AI工具复刻爆款视频广告仍需大量手动调整细节，单条视频耗时近一整天，批量生成多版本困难。

目标用户：效果营销人员、短视频广告投手、DTC品牌营销团队、视频广告代理公司、独立创作者。

证据：A single video will take an inter a whole day. ... With the same source file and 100 different run configurations, you can generate 100 different versions.

变现：开源核心编译器免费，提供付费云托管渲染、团队协作、模板市场、批量A/B测试和API调用；或按生成视频时长/渲染次数订阅收费。

MVP：先做一个Web编辑器，让用户粘贴脚本和参考广告，自动生成.scam描述文件，调用现有视频生成/渲染API输出带字幕、画中画和简单转场的3个版本，支持一键导出和基础版本管理。

竞品：Seedance 2.0、Kling

摘要：作者发布开源视频描述语言与编译器Hyper，用于把爆款广告复刻流程从手工一天压缩到分钟级，并支持同一源文件生成100个变体。帖子指出AI视频工具仍很难跨完整视频控制细节，但未给出付费意愿或收入数据。