Grok Bot automates talking-head editing with three prompts
In a hands-on test, author Meng Jian used three prompts to have Grok Bot analyze a local talking-head recording, identify pauses, mistakes, and repetitions, and produce a 9:16 captioned video without modifying the source file. He reported that the first edit removed about 15.05 seconds, followed by a style revision that enlarged subtitles, changed the title color, and generated a cover image and publishing copy. The article also includes the prompts and compares Grok Bot with the author's own 字流 tool, so the workflow and comparison are based on one user's test.
The report provides a concrete, reproducible example of an AI agent operating local media tools to complete multi-step video editing and asset creation through natural-language instructions.
大家好,我是孟健。
一个人做内容,最烦的就是录完口播坐在电脑前一刀刀切气口、转竖屏、配字幕。今天实测用 3 段提示词让 Grok Bot 在本地全自动剪掉口播原片的停顿口误、转成 9:16 竖屏带字幕,并顺带生成封面图与发布文案。
先看一眼最终调完的成片效果:

整个过程我只发了 3 段提示词,连原片文件都没碰一下。下面按我实际操作的步骤,把整套提示词和踩坑改样式的过程完整讲一遍。
准备工作:先让软件跑起来
Grok Bot 可以在官网(https://docs.x.ai/grok-bot/get-started)下载对应系统的安装包,打开后用符合条件的付费账号(比如 Premium 计划或绑定的账号)登录即可。