📡 Idea Radar 每天 8:00 / 18:00 推送 · 18:00 生成精选
reddit · SideProject · 69 分 ← 返回列表

I built an AI video detector over 3 months — looking for honest feedback

I've spent the last three months building Crackly AI, a tool designed to help people investigate whether a video is real or AI-generated. I wanted to work on a practical problem with AI rather than build another generic chatbot. I worked with a model around the 1-billion-parameter scale and built a product around the challenge of analyzing suspicious videos. I've now reached my first ₹10,000 in revenue. It's a small milestone, but after investing my own time and money into the project, it

五维评分

付费意愿
75
需求真实度
72
个人可落地
58
竞争空间
48
新鲜度
85

AI 分析

痛点:人们难以判断一段视频是否由 AI 生成,尤其在虚假信息、诈骗和版权纠纷场景中需要可解释、可信赖的视频鉴伪工具。

目标用户:内容审核团队、媒体与事实核查机构、法律/保险取证人员、社交平台、以及普通用户和中小企业。

证据:I've now reached my first ₹10,000 in revenue. It's a small milestone, but after investing my own time and money into the project, it feels meaningful.

变现:按次检测付费 + 订阅制(个人/团队套餐)+ API 调用计费;面向媒体、法务、保险、平台审核提供 B2B 报告与批量检测。

MVP:2 周内做一个网页端 MVP:上传短视频或粘贴链接,调用开源 deepfake/AI 视频检测模型或第三方 API,输出置信度、可疑片段和简短解释,并加入按次付费/邮箱收集付费意向。

竞品:Hive AI Video Detection、Reality Defender、Deepware Scanner、Optical AI、AI or Not、Illuminarty

摘要:作者用 3 个月做出 Crackly AI 视频鉴伪工具,并已获得首笔 ₹10,000 收入,说明有人愿意为视频真假检测付费。产品解决 AI 视频泛滥下的可信鉴伪需求,但该领域已有 Hive、Reality Defender 等竞品,独立开发者需在可解释性、垂直场景或低价按次检测上找差异。整体是值得关注的真实变现机会,但技术壁垒和竞争压力都不低。