📡 Idea Radar 每天 8:00 / 18:00 推送 · 18:00 生成精选
reddit · SideProject · 47 分 ← 返回列表

editing was the part of making reels i kept skipping, so i made a mac app that does the first cut for me

i post short reels on the side (instagram @stormej.studio). filming was fine. editing was where every reel died: picking takes, cutting the ums, redoing captions every time i trimmed something. so i built umber . you drop the clips you shot into a chat, it edits them into a reel with captions, and you change it by asking: "make it shorter", "start with the laptop shot". it runs on your mac and uses the claude or chatgpt plan you already pay for, so there's no separate ai bill. how it wor

五维评分

付费意愿
35
需求真实度
55
个人可落地
58
竞争空间
28
新鲜度
68

AI 分析

痛点:口播类短视频创作者卡在剪辑环节:挑片段、删口癖、每次重剪都要重做字幕,繁琐到让更新断更。

目标用户:在 Instagram/TikTok/Reels 上做口播或产品短片、有 Mac 但不会或不愿花时间剪辑的独立创作者与小工作室

证据:editing was where every reel died: picking takes, cutting the ums, redoing captions every time i trimmed something. / it uses the claude or chatgpt plan you already pay for, so there's no separate ai bill. / it's in a small private beta... i set each person up on a short call for now.

变现:买断制(本地运行、无推理成本,适合一次性 39-79 美元)或低价年订阅;增值项为云端渲染额度、多机位/长视频、字幕样式模板包

MVP:只做单机位口播这一条路径:本地 whisper 转录带词级时间戳 → 让用户已有的 LLM 返回「保留区间 / 去除口癖与废片 / 字幕文本」的 JSON → 代码校验切点对齐真实词边界 → ffmpeg 粗剪并按词时间戳烧字幕。不做 b-roll、不做花字动效,允许用户把剪切方案导回 CapCut/Premiere 继续精修。

竞品:Opus Clip、Descript、CapCut、Submagic、Klap、VEED

摘要:作者为 Instagram 侧项目创作者,因剪辑反复劝退而自建 Mac 应用 Umber:把素材丢进聊天框,用已付费的 Claude/ChatGPT 返回 JSON 剪切方案(保留区间、b-roll、屏幕文字、去口癖),ffmpeg 执行,字幕按同一词时间戳烧录以避免重剪错位。目前仅小范围内测,需 Apple Silicon Mac 与付费 AI 订阅,靠一对一通话开通,未披露定价、用户量或收入数据。