From 10 RMB to 1 RMB: How One Developer Slashed AI Video Costs with Codex + Obsidian

1. I Chased the Traffic of 'Life Scenario' Videos

Recently, 'Life Scenario' videos have been gaining huge traction on video platforms, often racking up tens of thousands of likes. Many in the community have been urging me to create a template for this. After analyzing a few high-traffic videos, I found the structure is simple: opening + theme freeze frame + main content.

But I've been busy with my Xiaohongshu virtual materials content. After dedicated refinement, my Xiaohongshu notes now average six-figure views per post.

Unable to resist the community's push, I squeezed in a few days this week to work on this Life Scenario project. At first, I shamelessly asked a friend for a template. They sent me a Coze template. I thought, great, I don't need to build from scratch. But when I ran my friend's template, the cost was over 10 RMB per run.

Switching to a Skill would be more cost-effective. Although the Skill's CapCut plugin isn't as rich, the cost-performance ratio is key. So I made the switch. After two days of work, I had the finished video:

The cost per video is between 1 and 3 RMB, a tenfold reduction. More importantly, changing the visual style is easy, unlike Coze where one change affects everything. I just send the topic and script:

After a few minutes, Codex generates the CapCut draft. My example here is too long; in practice, it should be cut in half.

2. Technical Breakdown

This Life Scenario Skill consists of two parts. The first part generates the script using my Obsidian knowledge base + Skill. The second part is the video generation Skill. I'll break them down in two subsections.

2.1 Life Scenario Script Skill

As someone who has been using AI for self-media for a year, I've found that Skill + Obsidian currently yields the best results for video or article scripts, far better than hard-tuning prompts. More importantly, as you accumulate more material in your Obsidian knowledge base, your scripts will keep improving. Using Obsidian is simple. For beginners, check out this article: How a One-Person Company Builds a Long-Term Memory System with WorkBuddy + Obsidian

To add script materials, I use a simple command: [Video Script] Put this script into the video script directory in the Obsidian knowledge base

Creating the Life Scenario Script Skill is straightforward. In Codex, input: Create a Skill: Refer to the scripts in F:\soft2\obsidian-vault\obsidian-vault\video-scripts\ (change to your knowledge base path), imitate the style, and when the user inputs a topic, generate the corresponding script.

2.2 Life Scenario Video Skill

The structure of a Life Scenario video is quite fixed:

Opening video + fixed voiceover: "Today's life scenario is..."

Theme freeze frame: Display the user's input topic, e.g., "Full-time civil service exam for 8 years, kicked out by parents"

Main content: Voiceover + subtitles + images/video clips + background music

So when I built the video Skill, I solidified this workflow. The user only needs to provide the topic and script; Codex handles the rest automatically.

Once this pipeline was running, Life Scenario videos became a semi-automated assembly line:

Topic + Script → Voiceover → Subtitle Alignment → Semantic Scene Segmentation → Image Generation → Partial Image-to-Video → CapCut Draft

The key is lower cost and higher controllability. Changing the visual style is as simple as a single command, unlike previous workflows that required various adjustments.

That's the complete breakdown tutorial for this Skill. This Skill has been added to the co-learning group's resource library; group members can directly use it.

3. Comparison: Coze vs. Skill for Video Production

Comparing the Coze and Skill versions for Life Scenario, I think each tech stack has its pros and cons:

Coze has a rich CapCut plugin ecosystem, but costs are slightly higher, node layouts are complex, and workflow creation requires dragging, dropping, and configuration.

Skill has low cost and high flexibility, but CapCut support is limited, requiring some manual work (e.g., adding CapCut effects manually). Skill creation is simpler—just tell it the logic of your workflow.

Which one to use is up to you. I haven't abandoned Coze because it has irreplaceable advantages, like convenient video production and a rich plugin ecosystem. For another domain I work on—comics—Coze remains a good choice. Using both tools together is also viable. Our ultimate goal in content creation is to achieve good results; maximizing the strengths of each tool is the optimal solution.

4. AI Agent Learning: How to Choose Tools

Many readers ask me: Coze or Skill? How to choose? They aren't mutually exclusive. If you have the energy, I recommend learning both. The core is flowchart thinking, a concept I've emphasized in previous articles. Just like today's Life Scenario breakdown—starting from the surface, breaking down the components, then the workflow—it's a reverse flowchart thinking process. Practice this more; the more videos you break down, the smoother it becomes.

That's all for today. I plan to create a dedicated account for this Life Scenario Skill and maybe even join the partner program~