A Director Skill Orchestrates Codex, HyperFrames, and Remotion Into an Auto-Editing Pipeline That Gained 1,000 Followers
theme: fancy
Hello everyone, I'm Xiaofeichang, a former programmer working hard to run a one-person company. Today I'm bringing you some of my experience with Codex automatic video editing. This workflow helped me gain over a thousand followers in the short-video space. I'm sharing the core prompt flow today as a reference for anyone who wants to try automatic editing.
1. My Automatic Editing Journey
I've been researching the automatic editing workflow for a long time. Two months ago, I started using HyperFrames to assist with video production. After making a few videos, my account got limited — the platform's feedback cited template homogeneity. I have to admit, the initial template was quite simple: just use a rough-cut skill to process the voiceover and add some animations to the background. Related skill guide: I Can't Make Videos, but I Used Codex to Run Voiceover + Auto Editing and Got 20+ Leads
Later I revamped it into IP animation, and the traffic improved. The IP animation idea was derived from my IP illustration work. At the time, many people liked my illustrations, so I thought, why not take it a step further and make them move? The result was very well received. Both IP animation and illustration are powerful buffs that strengthen your personal IP. Many followers told me that as soon as they see this little girl and the cat, they automatically recognize it as 'Xiaofeichang.' Related article: Building a Personal IP Without Appearing on Camera — I Made a Skill for One-Click IP Animation Videos
Recently I made another upgrade, combining IP animation, voiceover, and AI animation. The view counts are decent. Essentially, the current template integrates my previous skills, orchestrated by a director skill. The overall quality has improved a notch — at least it no longer looks like a slideshow. I'm still optimizing, of course, hoping to one day produce cinematic-quality videos. The road is long and hard...
2. Automatic Editing Technical Architecture
My current automatic editing workflow uses a director skill as the master controller to orchestrate other existing tools. Both the personal IP animation skill and the rough-cut skill are skills I built earlier. The director skill is responsible for connecting these two tools with the open-source HyperFrames and Remotion to produce the video. Just input a voiceover script and materials, and the director skill will organize and schedule the video production.
Creating this director skill prompt is also very easy. You can create it directly based on my prompt. Its core function is to chain the workflow. You need to ensure you have the rough-cut skill, the IP animation skill, and the two plugins HyperFrames and Remotion installed locally (if you don't have these two, have Codex install them first). Then it's about debugging the editing template. You can capture a clip from a good-looking video online and have Codex imitate it:
Debugging your own editing template is a long process. I'll expand on this tutorial later.
3. Some Recent Reflections
Recently, while running the Xiaohongshu project, group members have been reporting sales successes one after another, which makes me very happy. This has strengthened my determination to continue implementing AI in practice. I plan to do a mini-program business closed loop next, and AI-driven live selling after that. Compared to last year's pure AI tool tutorials, this year is more tiring but also more rewarding. By focusing on practical implementation, follower loyalty is stronger. Unlike last year's workflows, where competitors would take my workflow and immediately rebrand it as their own project, this year, the tools account for 10%, and my personal experience accounts for 90%. Followers recognize my personal IP, not just the skill.
I recently saw news that many AI bloggers have stopped updating. There was a period last year when I also wanted to stop. Back then, being an AI blogger mainly meant chasing hot topics — replicating whatever was trending with workflows, without any practical implementation. That approach was wrong: exhausting, hard to break through, difficult to monetize, and easy to become a tool-user. Later I changed my strategy, focusing on practical monetization. From last year's WeChat public account comics to this year's Xiaohongshu virtual materials training camp, I implemented everything myself and then helped others succeed. The results prove my method is correct. Many followers have told me I'm a highly implementation-focused AI blogger. I've successfully imprinted this label in my followers' minds — I'm no longer just an AI tool person. That's all for today's article. Thanks for watching~