Overview
When you say "I want to make video with AI," you might mean two opposite things. The first is editing: cutting, cleaning up, and captioning footage you already shot. The second is generating: creating images that were never filmed. The market calls both "AI video," and that's why you search for the wrong tool and conclude the technology doesn't work.
This chapter gives you a 5-second test to tell the two apart. You won't need a subscription for this. You'll open CapCut, watch AI work on your own video, and understand what it does and what it refuses to do. When you're done, you'll never confuse "AI doesn't work" with "I brought the wrong problem."
Key Concepts
Assisted editing is AI that acts on what already exists. It takes your raw video and cuts silences, removes mistakes, transcribes speech, suggests cuts. It doesn't create a single new frame. CapCut, for example, has automatic captions: the AI listens to the audio and writes the text. If you said "bread," it writes "bread." If you said nothing, it doesn't invent speech. Its limit is your material. And it doesn't decide for you: it speeds up the grunt work, but the judgment of what stays is yours.
Generation is another family. It takes text, an image, or a reference video and produces new pixels. A prompt like "blue hamburger falling in slow motion" becomes a few-second clip that never existed. Tools like Runway, Luma, Pika, Sora, and Veo do this. The limit is the opposite: they don't organize your material. You don't feed them 40 minutes of footage and get a finished video. The result rarely comes out perfect on the first try; you repeat the prompt until you get close to what you imagined.
The mental model is short: if the problem is "I have too much material and don't know what to cut," you need editing. If the problem is "I don't have the image and can't film it," you need generation. There are tools that do both, but they're different products, with different prices and limits. That's why the question "what's the best AI video tool?" has no single answer. The answer depends on the verb: cutting or creating.
Execution Flow
- Open CapCut on your phone and tap "New project." Import a clip of at least 1 minute that you shot with the camera — it can be you talking to the front camera. The screen shows the timeline at the bottom and the video on top. CapCut has a free plan; some features require a subscription, and the price shows on screen when you try to use them.
- Tap "Captions" in the toolbar and choose "Auto captions." Select Portuguese. The AI transcribes the audio and places the captions on the timeline. If the audio is too quiet, it warns you it couldn't recognize the speech. It doesn't invent the text.
- Notice what happened. The captions appear as blocks on the timeline, aligned with your voice. No frame of the video was altered. The AI edited the text, not the visual content. That's the typical behavior of assisted editing.
- Try "Remove silences," if it appears in your menu, or manually cut the parts where you're not speaking. CapCut removes the empty gaps and joins the parts with speech. The video gets shorter, but every image on screen came from your original recording.
- Compare with a generator. Open a generator's site with a free plan, like Luma, type "a cat skateboarding" in the prompt field, and click generate. If the field is empty, the button is disabled — it requires text. The result is a short clip, created from scratch, that doesn't exist on your phone. One cut what you had; the other created what you didn't have.
Applied Scenarios
Scenario 1: you recorded 40 minutes of a homemade bread class, with long pauses, coughs, and a mistake in the middle. You open CapCut and expect "AI video" to solve everything on its own. It won't. Editing AI cuts silences and transcribes, but you're the one who decides what stays. After 20 minutes of work, you have 12 minutes of video with captions. Before: 40 minutes nobody watches. After: 12 minutes you can publish. No scene was created — the AI reorganized what existed.
Scenario 2: you want to show a "blue hamburger falling in slow motion" and have no way to film it. You try with your phone and the result is ridiculous. Then you realize the problem isn't editing, it's generation. You write the prompt in the generator, wait a few minutes, and get a few-second clip. It's not perfect: the cheese melts in a weird way, the background changes color. You repeat the prompt, swap "falling" for "bouncing," and the result improves. Before: an idea you couldn't show. After: an imperfect clip that communicates the idea.
Common Mistakes
- Recording 40 minutes and asking the AI to "make a video." It cuts, cleans up, and adds captions, but the decision of what stays is yours.
- Thinking the generator will edit your footage. It creates new scenes, it doesn't organize yours.
- Giving up on editing because "I don't know how to edit." CapCut automates silence removal and captions, but you still choose what stays.
- Using a generator for content that needs precision, like a recipe. It invents details that don't exist.
- Signing up for a paid plan before testing the free one. Most generators have a free version with a limit on seconds or credits; the paid plan price shows up on the upgrade screen.
Pro Tip: Before opening any tool, write in one sentence what you want. If the sentence starts with "cut," "organize," or "clean up," it's editing — go to CapCut. If it starts with "create," "show something that doesn't exist," or "imagine," it's generation — go to a generator. This 5-second test saves you hours of using the wrong tool. Today's top-tier generators produce short clips, a few seconds at a time; that's normal behavior, not a flaw in the tool.
Practical Exercise
Grab your phone and record 2 minutes of yourself talking about your day. It doesn't need to look good. Open CapCut, import the clip, and generate automatic captions. Then remove the silences or cut the parts without speech. Export the result. Done criteria: you have a video under 1 minute, with captions, and you can answer in one sentence whether the AI created any new frame. If the answer is "no," you understand the difference. Now, write down a prompt for a generator on paper, like "a hot loaf of bread coming out of the oven in slow motion." You don't need to generate it yet — just write it down. Chapter 2 will show you where to use that prompt.
Implementation Checklist
- I can explain the difference between editing and generating using an example of my own.
- I know where the automatic captions feature is in CapCut and what to expect from it.
- I can classify a task of mine as editing or generating before choosing the tool.
- I know that generators create short, imperfect clips, and that's normal.
- I don't sign up for any paid plan before testing the free one.
Chapter Summary
- Assisted editing reorganizes what exists; generation creates what doesn't.
- CapCut is an example of editing: it transcribes, cuts, adds captions, but doesn't invent frames.
- Generators like Luma and Runway create short clips from a prompt.
- The most common mistake is taking the wrong problem to the right tool.
- In chapter 2, you'll learn the four ways to generate video: text→video, image→video, video→video, and avatar. Then you'll know which door to open for each idea.
---