I Let AI Edit an Entire Reel for ₹50: ChatGPT vs Antigravity vs Claude

I got AI to edit an entire reel for me, and it cost around ₹50.

I do not mean cuts and captions. I mean motion graphics, sound effects, visuals over the footage. The whole editing job, the part I used to either do badly myself or put off for a week.

The three I tried#

I gave the same kind of job to three tools:

• ChatGPT
• Google Antigravity
• Claude

And honestly, in my experience, Claude is by far the best at this. ChatGPT and Antigravity were more or less on a similar level. Claude was noticeably ahead.

This is one person's experience over a handful of reels, not a lab benchmark. The tools change every few weeks, and your footage is not my footage. But the gap was big enough that I stopped switching between them.

How it actually works#

Here is what makes this interesting: you do not need to know every editing technique.

The workflow is closer to briefing an editor than to using editing software:

1. Give it the raw footage. The unedited clip, straight off the phone.
2. Explain your thoughts out loud. Talk it through: what the video is about, where the energy should pick up, what you want the viewer to feel at each point.
3. Say exactly what you are imagining. Something like "when I say the price, I want the number to slam onto the screen" or "put a quick whoosh on that cut". The more specific the picture in your head, the better the first draft.
4. Keep refining. The first version is never the final one. You watch it, you say what is off, it fixes that. A few rounds usually gets me there.

You are doing the directing. The AI is doing the hands-on work. That is a real shift for anyone who has ideas for how a video should look but has never learned a timeline editor properly. Which, until recently, was me.

Where the ₹50 comes from#

I calculated the cost per reel and it came to ₹40 to ₹50, including the editing and the sound effects.

That number is usage: what the AI work on one reel came to when I added it up. It will be different for you. A longer video, more rounds of changes or a bigger model will all push it up.

For comparison, I ran a separate test where I had three Claude models each make a 30-second music video and logged every rupee: Haiku vs Sonnet vs Opus. The short version is that the model you pick moves the bill a lot more than you would expect.

What it is good at, and what it is not#

Good at:

• Motion graphics and text that hits on the beat.
• Sound effects in the right places, which is the single cheapest way to make a reel feel edited.
• Turning a vague idea ("make this bit feel bigger") into something concrete you can react to.
• Not getting bored on the fourth round of changes.

Not good at:

• Taste. It will happily give you too much. More effects, more movement, more everything. Knowing when to say "less" is still your job.
• Knowing your audience. It does not know which joke your followers will get.
• Fixing bad footage. A boring clip with great graphics is a boring clip with great graphics.

Should you try it?#

If you are a content creator who likes experimenting, this is a pretty crazy time to be creating. The cost of trying an idea has dropped to roughly the price of a samosa and a chai.

My advice is to start with a reel you were not going to make anyway. One where it does not matter if the result is strange. Talk the idea through, see what comes back, and spend your energy on the refining rounds. That is where the result goes from "AI made this" to "this is mine".

If you want a tool that does one specific version of this, turning a project you just built into a launch video, I wrote about that too: Brag.