CapCut Made Simple: From Beginner to AI Skill for Your First Viral Video

CapCut (JianYing) is a free editing software from ByteDance. For newcomers learning video editing, this is the best choice if you're not a professional editor.

Desktop versions are available for both Windows and macOS. Mobile, web, and desktop drafts can sync with each other.

It's free, has a Chinese interface, and offers many templates. It has the lowest barrier to entry for making your first video.

But it has an unspoken problem—the software itself isn't hard; the difficulty is knowing what to click first when you open it.

This article is split into two parts.

Part 1 covers how to use CapCut itself. From installation to export, everything is done manually to cut out a publishable video.

Part 2 introduces an open-source skill. You describe your needs in plain language to an AI, and it builds the CapCut draft directly—subtitles, voiceover, and music all set up at once.

The two parts build on each other. Only after walking through Part 1 will you understand exactly what Part 2 saves you.

Table of Contents

Part 1 · Manual CapCut Operations

• 1. Which Version & Where to Install — How to choose between three versions, why you must note the draft directory first
• 2. Recognize Four Areas — Media panel, preview window, timeline, parameter panel
• 3. First Cut: Getting Footage onto the Timeline — Import, split, delete, reorder
• 4. Voiceover: Text-to-Speech — Type text to generate narration, easy to revise
• 5. Subtitles: Auto-Caption Recognition — Recognize, fix typos, adjust phrasing, dialects
• 6. Music & Volume — Don't let voice and music compete
• 7. Export — How to set resolution, frame rate, bitrate

Part 2 · Let AI Operate CapCut Directly

• 8. What This Skill Is — Not a replacement, but an operator sitting in front of CapCut
• 9. Installation — One-click script, directories for four editors
• 10. Run Your First Video — Verify environment, run tests, handle missing drafts
• 11. What It Can Do — Six scenarios, each with copy-paste-ready prompts
• 12. What It Can't Do — Tasks beyond its reach
• Common Pitfalls — Six places where things get stuck
• Quick Reference — Ten ready-to-use commands

Part 1 CapCut Basics: Manually Edit Your First Video

1. Which Version & Where to Install

CapCut has three versions: mobile, web (offline since 2024), and desktop Pro.

Cloud drafts sync between two versions—you can edit halfway on your phone and continue on your computer.

This article uses the Desktop Pro version. Two reasons: First, multi-track editing, subtitles, and export parameters are easier to handle on desktop. Second, the skill in Part 2 only recognizes the desktop version.

Download from the official site: https://www.jianying.com/ . The desktop version is called JianYingPro.

The international version of CapCut is just "CapCut." CapCut is made by the same team as an international edition—the interface looks similar, but draft formats aren't compatible. The skill in Part 2 doesn't support it at all.

After installing, do one thing first: Note down the default draft directory path.

1Windows:
2C:\Users\Administrator\AppData\Local\JianyingPro\User Data\Projects\com.lveditor.draft
3
4macOS:
5/Users/<username>/Movies/JianyingPro/User Data/Projects/com.lveditor.draft

All projects, media assets, and subtitle caches live in this directory. Finding drafts manually requires it, and having AI generate drafts also requires it.

If you changed the installation location during setup, recheck this path—it's the most common error point in Part 2's environment check step.

The simplest way to edit in CapCut is using templates—replace images or text assets, then export to produce a finished video!

2. Recognize Four Areas

Open CapCut Pro, click Start Creating on the home screen to enter the creation module. The interface has four areas.

• Top-left: Media panel. Contains media, audio, text, stickers, effects, transitions, filters, adjustments
• Top-center: Preview window. Playback controls in bottom-right corner, supports fullscreen
• Bottom: Timeline. All editing actions happen here
• Top-right: Parameter panel. Shows adjustable properties for whatever is selected

The timeline is the main battlefield.

Tracks come in two types.

Video tracks can stack multiple layers—the upper track covers the lower one. Picture-in-picture and subtitle backgrounds rely on this.

Audio tracks also support multiple layers—music, narration, and sound effects each occupy their own track without interfering.

Editing boils down to three things: put footage on tracks, cut out unwanted parts, arrange the rest in order.

3. First Cut: Getting Footage onto the Timeline

1. Open CapCut, click "Start Creating" to make a new draft
2. Click "Import" in the media panel, drag videos, images, audio in
3. Drag assets from the media panel to the main video track on the timeline

Once footage is on the track, next comes cutting.

The white line is the playhead. Drag it to the position where you want to cut, select the clip, press Ctrl + B to split. On Mac, use Command + B.

After splitting, select the unwanted segment and press Delete to remove it. If you cut wrong, Ctrl + Z to undo.

Reorder by dragging. To change duration, drag the ends of clips.

CapCut auto-saves, so don't worry about saving.

The most common beginner mistake is cutting too finely. Three-second cuts, five-second cuts—viewers haven't even seen the frame before it jumps away. Rough-cut the whole video first to remove bad shots, then review pacing afterward.

4. Voiceover: Text-to-Speech

This is one of CapCut's most valuable features. You type text, and it reads it aloud.

1. Click "Text" at the top → "New Text"
2. Paste narration sentences one by one
3. Select the text, find "Text-to-Speech" in the right-side parameter panel
4. Choose a voice, click Generate

Generated TTS audio lands on a separate audio track, aligned with the text block positions.

Commonly used voices in the library:

What makes AI voiceover truly convenient isn't speed—it's that revisions don't hurt. Mispronounced word, wrong tone? Fix that line and regenerate. With real human recording, you'd have to sit back down at the mic.

Sixty seconds of narration—typing plus generation—can be done during your commute.

5. Subtitles: Auto-Caption Recognition

Top bar "Text" → "Smart Subtitles" → "Start Recognition."

Select language, wait while it processes online, subtitles land on the subtitle track—one block per sentence, synced with audio.

After recognition, do two things.

First, read through and fix typos. Double-click any block to edit directly. If a section has many errors, right-click that block and choose "Re-recognize."

Most misrecognized words fall into three categories: names, technical terms, mixed Chinese-English. Manually check these three types; the rest is mostly usable.

Second, adjust phrasing. Next to the recognize button there's a dropdown—you can turn off "Smart Phrasing" and switch to breaking by speech pauses. Talking-head videos sound more natural with the latter.

Dialects are supported too. Currently: Minnan, Shanghainese, Central Plains Mandarin, Southwestern Mandarin. Content with non-standard Mandarin actually gets better accuracy using the corresponding dialect model.

Version differences matter: Newer versions moved subtitle recognition behind membership paywall. Entry points and permissions depend on your installed version.

6. Music & Volume

Click "Audio" → "Music Library" in the media panel. The library is categorized by scene, mood, instrument—preview and drag suitable tracks to the audio track.

Select the music block, adjust volume and add fade-in/fade-out in the right panel.

One rule of thumb: Narration is the star—push music down until voice is clearly audible.

The most common beginner flaw is voice and music fighting over volume.

Easy fix: Set music block volume around 20%, drop it another notch during speaking segments.

7. Export

Click "Export" in the top-right corner.

Three parameters: Resolution, Frame Rate, Bitrate.

Max resolution 4K, max frame rate 60fps, bitrate has three tiers.

For Douyin, WeChat Channels, Bilibili—1080P + 30fps with recommended bitrate is sufficient. Smaller file, faster upload, platforms compress again anyway—exporting 4K is wasted effort.

When you need to export just one segment, you can limit the range during export.

That completes Part 1. You can now publish your first video.

Part 2 Let AI Operate CapCut Directly

8. What This Skill Is

In one sentence: It's not a CapCut replacement—it's an operator sitting in front of CapCut.

You write what you want in plain language, and AI builds the draft—asset arrangement, subtitles, voiceover, music, transitions—all configured at once. Then open CapCut, and the draft is already waiting.

Project name: jianying-editor-skill, MIT licensed open source, 3.7k stars, 481 forks, 117 commits. Latest version v1.7, updated September 11, 2026.

Built on the open-source library pyJianYingDraft. This skill does two things on top of it: adapts to CapCut 5.9+ new draft structure, and wraps an AI-facing call interface.

Compatible with Antigravity, Trae, Claude Code, Cursor—four types of AI editors.

9. Installation

Three prerequisites: CapCut Pro desktop installed, an AI editor capable of running skills, Python on your machine.

Easiest install—throw it at your agent:

1Help me install this skill https://github.com/luoluoluo22/jianying-editor-skill.git

Windows has a one-click script, run in PowerShell:

1irm is.gd/rpb65M | iex

Manual install via git clone, choose directory based on your editor:

1git clone https://github.com/luoluoluo22/jianying-editor-skill.git
2
3# Claude Code          .claude/skills/jianying-editor
4
5# Cursor / VSCode / Generic       skills/jianying-editor
6
7# Trae                           .trae/skills/jianying-editor
8
9# Antigravity / Gemini Code Assist      .agent/skills/jianying-editor

Then install dependencies:

1pip install -r requirements.txt
2playwright install chromium
3
4# Optional
5# Install Python and FFmpeg
6brew install python ffmpeg
7
8# Create and activate isolated Python environment
9python3 -m venv .venv
10source .venv/bin/activate
11
12# Install Edge TTS and Pillow
13python -m pip install edge-tts pillow
14
15# Verify installation
16python --version
17ffmpeg -version
18python -c "import edge_tts, PIL; print('Edge TTS and Pillow installed successfully')"

The playwright line is for screen recording and web-to-video conversion.

For future updates, just git pull in the skill directory. So don't write editing scripts inside the skill directory—write them in the project root, otherwise git pull conflicts with local changes.

10. Run Your First Video

After installing, verify the environment first. Tell the AI:

Check if the CapCut automation environment is working properly

It runs environment validation, checking three things: Is FFmpeg present? Are Python dependencies complete? Was the CapCut draft directory found?

The third is most error-prone. If draft directory detection fails, tell the AI the actual path you noted in Part 1.

Once environment passes, run a test:

Just cut a random test video for me to see

It creates a new draft with demo assets and test subtitles.

After running, open CapCut—not seeing the new draft is normal.

CapCut doesn't refresh the draft list in real time. Restart CapCut, or enter any old draft and exit—the list updates. This is the most frequently asked pitfall.

Practical effect video after installing the skill

Effect is minimal, just to verify the skill works

Video prompt below:

1Use jianying-editor skill to create a ~30 second classical poetry recitation video:
2
3Work: Nalan Xingde's "Mu Lan Hua · Imitating Ancient Style Farewell Poem"
4Text:
5If life were always like first meeting,
6Why would autumn wind grieve painted fans?
7Casually hearts of old friends change,
8Yet they say old friends' hearts easily shift.
9At Lishan, talk ended half through clear night,
10Rainy bells wept yet no regret remained.
11How could the unfaithful brocade-clad youth
12Ever promise wings joined, branches intertwined?
13
14Production requirements:
151. 1920×1080 landscape, 30fps, total duration ~30 seconds.
162. Use gentle, soothing, slightly melancholic Chinese female voice, natural reading with appropriate pauses between lines. Read only the body text, not title, author, or variant notes.
173. Auto-generate synchronized line-by-line subtitles, use clear Song font, avoid obscuring main visual elements.
184. Add soft guqin or guzheng instrumental music, volume lower than narration, fade in/out at start and end.
195. Use elegant rice-paper, ink-wash ancient style visuals, change imagery per line: first meeting, autumn wind painted fan, old friends parting, changing hearts, Lishan moonlit night, rain bells, brocade-clad wanderer, paired birds intertwined branches. Soft transitions between scenes with subtle motion.
206. Title and author shown as small concise text, body subtitles prominent.
217. Use my provided text version, don't add "variant" notes. Reference: https://www.guwendao.net/mingju/juv_43f5bfba7229.aspx
228. Complete production directly without item-by-item confirmation; if assets unavailable, use reasonable alternatives and explain upon delivery.
239. Deliver playable MP4, SRT subtitles, and editable CapCut draft. Even if Mac can't auto-export from CapCut, generate final MP4 via other available methods.
2410. Export final video per my Bilibili compatibility specs, filename includes "bilibili", verify with ffprobe and complete decode validation before delivery.
25
26Save files to independent directory under /Users/username/Documents/Codex/today_date/, provide clickable links to video and subtitles upon completion.

11. What It Can Do

Listed by scenario, each followed by prompts you can copy directly.

Vlogs & Basic Editing

Cut all videos in D:\Assets\ChengduTrip into a Vlog, add upbeat BGM, title it "Strolling Chengdu"

Stitch together the clips I just recorded, add fade transitions between each segment

Add a 3-second title text at the beginning of this video, use that "retro typewriter" animation effect

Script-to-Video

Write a science explainer script about "How to Stay Up Late Without Harming Your Health," pair with gentle female voiceover, auto-generate subtitles

This triggers its one-click pipeline: write script, generate voiceover, align subtitle track—all at once.

I have a recorded narrator.mp3, recognize subtitles, then auto-select matching visuals from asset library

This uses semantic asset matching. It analyzes visuals, narration audio, and subtitles together, picking shots by content—not stacking by filename order.

Make a two-person dialogue video: male voice says "Hello," female replies "Good morning," subtitles auto-switch style based on speaker

Screen Recording & Software Tutorials

Record an Excel operation tutorial, auto-add smart zoom animations after recording

Record a 10-second demo of http://localhost:3000 webpage, then import into new draft

Screen recording automatically inserts zoom keyframes at click locations. Tutorial videos rely on this to guide viewer attention along the cursor.

Asset Search

Search CapCut for "watercolor" style filters, list them for me

Apply "cinematic" filter to all clips in this video

Filters, effects, transitions in CapCut can be searched by name. Want an effect? Describe it, and it finds the closest match.

Project Management

List the 10 most recent CapCut drafts

Draft "Untitled Project" seems corrupted, won't open, help repair it

For corrupted drafts or version mismatches, it rebuilds a compliant project file using overwrite mode.

Export

Export "Final Version Video" project as 4K 60fps MP4, save to desktop

12. What It Can't Do

This section matters more than the previous one.

It only manages the timeline and draft layer. Rendering and preview still run within CapCut itself—it builds the draft, then clicks export for you.

Specifically cannot do:

• GPU-realtime-computed features can't be triggered. Smart background removal, beauty filters, speech recognition subtitles—it can't touch these
• CapCut's built-in AI buttons can't be clicked. "One-Click Video" and "Image-to-Video" require manual clicking
• CapCut International Edition unsupported. Only recognizes domestic JianYingPro
• Mobile not supported. Only handles desktop draft projects
• macOS can't auto-export

Platform support comparison:

Auto-export is Windows-only. It relies on Windows UI Automation to simulate mouse clicks, so it only works on Windows.

Also most stable on CapCut 5.9 or lower. When newer interfaces change, control positions shift, and auto-clicks may miss targets.

On Mac, take a step back—draft generation works normally, just manually click export at the end. This isn't a defect; UI automation coverage simply extends this far.

Common Pitfalls

Draft not showing CapCut doesn't refresh draft list. Restart CapCut, or enter an old draft and exit.

Export fails midway Don't touch mouse/keyboard during export. Script simulates operations—human interference disrupts it.

Mac export command no response By design. Mac doesn't support auto-export—must click manually.

After importing assets, error "Media lost detected, please relink before editing" Issue with CapCut 5.9+ versions, fixed in v1.7. Git pull to update first.

Editing scripts written in skill directory Write in project root under scripts/. Writing in skill directory causes git pull conflicts.

Auto-export resolution wrong Export resolution/frame rate follows CapCut's export settings, not what's written in script. Confirm CapCut's default tier first.

Quick Reference

Final Thoughts

Manual editing and AI editing aren't either/or choices.

For your first video, walk through manually first. Know which track subtitles land on, how voiceover generates, where export parameters are set—so when problems arise, you can locate them.

After completing this walkthrough, hand off to AI. What you save is arrangement and repetitive configuration time—not understanding cost.

CapCut itself is free, and that skill is MIT open source. Together, zero-cost starting point.

I'm Xiang Feng @xaiwind, a creator focused on going global and AI. #CapCut #VideoEditing

This article was created with AI assistance combined with personal practice. If helpful, please like, bookmark, follow.

My articles come with after-sales support—contact me or comment below. Progress a little every day, and we'll eventually reach our destination!

Appendix · References

• jianying-editor-skill (MIT open source): https://github.com/luoluoluo22/jianying-editor-skill
• Underlying dependency pyJianYingDraft, Apache License 2.0
• JianYingPro official site and version update notes

Previous Excellent Video Articles:

Screen Recording & Editing Software Recommendations

https://x.com/xaiwind/status/2097594816616116417