The days of you needing to learn the fundamentals of programming, an entire new language, or a full new framework just to build a simple idea are fading faster every day. While the claim "AI can code" didn't quite make sense back when ChatGPT was first launched, the promise has turned into something very real now. We have tools like Claude Code and OpenAI's Codex which are geared at developers who want AI as a coding partner, but then there's a different category entirely: tools built for people who just have an idea in their head and want to see it exist.

That's where Lovable, v0, and Bolt come in. All three of these tools have extremely simple UIs, more akin to a chat window than an IDE or a terminal window. You type what you want to create, and they begin bringing your idea to life. These tools also let you deploy the whole thing in-app, meaning you can go from a rough idea to a shareable live URL all without leaving your browser. That said, the question has gone from "should we use these tools" to "which of these tools actually does the job." I gave all three the exact same prompt to find out.

I wanted to build a photo booth app

The prompt that started it all

I've done a few of these app comparisons before, and I've found that the best way to put these tools to the test is by attempting to create something I genuinely want to. Not a generic to-do list app or a calculator — I think we've established that AI tools can handle those just fine. I wanted something with enough personality and specificity that the AI actually has to make decisions.

So, for this comparison specifically, I decided I'd build a vintage photo booth app. This idea stemmed from an actual conversation I was having with a friend about how those photo strip booths are everywhere right now, and how it'd be fun to have a browser version you could just pull up on your phone or laptop.

Now, I'm sure this isn't an entirely original idea, and there are probably a dozen of these already out there. The goal was simply to build something I genuinely wanted to, and see how it holds up when actually using it.

Along with building something I want to, a massive part of these experiments is keeping the prompt intentionally vague. I don't spell out every requirement of what I'm attempting to create, but I always provide enough detail to give the AI a clear direction. The rest is up to it. I want to see what decisions it makes on its own. What it adds that I didn't ask for, what it interprets from the vibe, and where it falls short.

Every time I mention this approach in an article, people point out that I'm intentionally sabotaging the final output. I want to make it very clear that I'm testing this from the perspective of someone who isn't a developer. I know the fundamentals of programming well, and I could in theory build this myself entirely. But for these tests, I put that aside. The whole selling point of these tools is that anyone can use them, so I test them the way a non-developer would. If I have to write a perfectly structured technical spec to get a decent result, that defeats the purpose.

Bolt simply disappointed me

But I can't blame the tool entirely

Out of all three tools, Bolt was the only one that couldn't deliver a fully working app. Lovable and v0 both managed to build something workable with just a single prompt, but Bolt couldn't even with six rounds of re-prompting. The most fundamental part of the entire app, the camera, never loaded. The page would either freeze, become unresponsive, or just sit there with a blank feed.

The more frustrating thing here was Bolt's limits. To be clear, I used the free version of all three tools, and Bolt was the only tool where I hit limits with just one project. I never even got to evaluate the actual photo booth, the themes, the strip layout, the download, because all my free tier usage was burned on debugging camera permissions (which never worked by the end of it). In fact, I had to spread out my testing across three days because I kept hitting the daily token cap mid-session. Every failed attempt and every follow-up prompt ate into my allowance.

For context, the free tier gives you a million tokens per month with a 300K daily cap. Bolt has paid tiers starting at $25/month that remove the daily cap and bump you up to 10 million tokens, but I shouldn't need to pay just to get a camera to turn on! In fact, the free plans are supposed to be the selling point. They're supposed to be what gets you hooked before you upgrade. If the free experience can't even produce a working prototype, why would I trust the paid version with anything more complex?

All that said, this did mean all I can judge Bolt's output on is the UI, and that too just the homepage. Credit where it's due, the UI was minimalist and matched the vibe I was going for — soft pastels, rounded corners, a clean layout. It understood the aesthetic. The problem was everything underneath it.

Lovable and v0 both delivered

But one felt more alive

Unlike Bolt, both Lovable and v0 actually shipped working photo booth apps from a single prompt. Camera loaded, countdown worked, photos captured, strip generated, download worked. Feature-for-feature, they were nearly identical. Both gave me pastel themes, a film grain toggle, a date stamp, a watermark, and a downloadable PNG as I had asked. I used the free tier on both, and neither one made me feel like I was hitting a wall.

The experience was smooth from prompt to finished product. The only gap I felt was that Lovable's output felt a bit more like a finished product.

The strip preview had a floating animation, the background was a gradient that shifted between pastels, and little details like the microcopy ("smile bestie...", "warming up the lens...") made the whole thing feel a lot more intentional. v0's output worked just as well functionally, but it just fell slightly flat in comparison. The layout was clean and minimal, the features were all there, but it just felt very static.

Lovable won, but only by a margin

If v0's output had just a bit more life in it, this would've been a straight tie. I want to reiterate again that I tried these tools out as a non-developer would. I could've talked about the models each tool uses under the hood, or how they handle version control, or the quality of the generated code. But that wasn't the point. The point was: I have an idea, I type it in, does it work?