How to Automate Professional Video Creation in CapCut with Codex App: A Complete Guide to VectCutAPI

We are used to artificial intelligence writing code, building websites and bots on a simple description. But what if you need to quickly turn a raw screen recording with a demonstration of the application into a ready-to-publish video? No long hours in CapCut, no manual effects and no timeline tinkering.

The solution is Codex App (from OpenAI) + VectCutAPI (open source from sun-guannan). This bundle allows you to fully automate video processing: add effects, text, animations, voiceover, music and cinematic color correction directly through a regular chat with an AI agent. The finished draft opens in CapCut with one click – then you can finalize manually or immediately send it to a render in the cloud.

## 1. How it works

• **Codex App is no longer just a chat. This is a full-fledged desktop application from OpenAI for working with AI agents. It can run MCP servers (Model Communication Protocol), work with Git, run local services and think like a senior video editor.
• VectCutAPI is an on-premises or cloud-based API that takes over the hard work of CapCut: creating drafts, managing keyframes, animating text, adding effects, subtitles, stickers and audio. Everything through JSON or MCP is perfect for AI agents.

The result: a raw two-minute screen-recording demo of your app turns into a ready-made vertical Reels/TikTok or horizontal YouTube video in just 5-10 minutes of conversation with the AI.

## 2. What you need in advance

### Definitely:

1. Codex App (macOS, free for ChatGPT Free/Go; Plus/Pro – higher limits). Downloaded from openai.com.
2. VectCutAPI (GitHub: https://github.com/sun-guannan/VectCutAPI).
3. CapCut (desktop or mobile version).
4. FFmpeg (installed in PATH).
5. Python 3.10+.

## 3. Step-by-step adjustment

### Step 1: Installation of VectCutAPI

git clone https://github.com/sun-guannan/VectCutAPI.git
cd VectCutAPI
python -m venv venv-capcut
source venv-capcut/bin/activate    # macOS/Linux
# venv-capcut\Scripts\activate     # Windows

pip install -r requirements.txt
pip install -r requirements-mcp.txt   # для MCP
cp config.json.example config.json

Edit config.json (insert paths to CapCut drafts and port 9001).

Launch the servers:

python capcut_server.py     # HTTP API
python mcp_server.py        # MCP (для Codex App / Claude)

Servers are ready. Now AI can address them.

### Step 2: Connect Codex App to VectCutAPI

1. Open the Codex App.
2. Create a new project or chat.
3. Write in the Prompt:

I configured VectCutAPI to localhost:9001 and MCP server. You're my AI Video Editor now. Use the MCP tool-calling for all CapCut operations. Our goal is to turn raw demo videos into premium promos

Codex App will automatically detect the MCP server and use VectCutAPI tools as built-in skills.

## 4. Main workflow: from raw video to finished video

### Example of task

You have a raw video - a screen demonstration of the functionality of the application (for example, VibeCode Wiki). You need to add cinematic transitions, animated text with benefit, glow effects on clicks, modern color correction, voiceover or music and the final CTA.

Prompt for Codex App:

You are an Apple-level world-class motion designer. I have a raw video [insert a link to a file or URL].

Task:
1. Create a new draft 1080x1920 (9:16 for Reels) or 1920x1080 (16:9).
2. Import the video as the main track.
3. Put down dead time, tighten the timing.
4. Add cinematic color grade: deep blues + cyan accents + subtle glow.
5. For every important tap/click, ripple effect + animated outline.
6. Add text (Inter/SF Pro style): feature name + 1-second benefit. Animation: fade + scale + glow.
7. Add upbeat tech music (volume 25%) + voiceover in Russian (confident male voice): "This is how it works..."
8. At the end - logo animation + CTA "Try for free" + link to vibecode.morecil.ru.
9. Save the draft and link to the dfd * folder.

Use only VectCutAPI MCP tools. If necessary, first analyze the length of the video.

Codex App will call the required commands (create_draft, add_video, add_text, add_video_keyframe, save_draft). In 30-90 seconds, you will have the draft folder ready. Copy it into the CapCut drafts folder - the project will open with effects.

## 5. Advanced opportunities and nuances

Important shees:

• Long video (>60 seconds) → break into segments.
• Add to the prompt: “Preserve all on-screen UI text exactly.”.
• Codex App can first do upscaling via FFmpeg.
• Local vs cloud render → for speed use local CapCut.

References:

• VectCutAPI: https://github.com/sun-guannan/VectCutAPI
• Codex App → openai.com/codex-app