I Launched a SaaS, Forgot About It, Then Discovered People Were Paying for a Broken Product

Marcus — Thu, 12 Feb 2026 18:30:00 +0000

When I started making music videos for Youtube, I kept running into the same problem: manually timing lyrics took forever.

Upload song. Listen on repeat. Note when each line starts. Export subtitle file. Adjust timing. Repeat.

At least an hour per video. After doing this dozens of times, I thought there has to be a better way.

So I built one. Then I forgot about it.

The Accidental Launch

I spent a few weeks building an MVP. Bought a domain. Deployed it. Then moved on to a different idea.

Over the next few months I'd be getting some emails which I assumed were just recurring charges from when I subscribed to the app myself.

Another month passed. I logged into Stripe to cancel my subscription.

There were four customers.

Four people had found my website, signed up, and paid. I hadn't submitted it to search engines. I hadn't announced it anywhere. I barely remembered it existed.

Then I opened the app and tried using it. Parts of it were broken. Some features didn't work. The UI was pretty horrible.

I was embarrassed. Refunded everyone immediately and sent apology emails and bonus transcription minutes (according to Convex none of them have logged in and used the minutes, I wonder why).

But something stuck with me: if four people found my broken tool and paid for it, maybe there was something here worth fixing.

So I rebuilt it properly. That was late December 2025.

The Problem

Creating synced lyrics is surprisingly painful.

Content creators need subtitle files for TikTok and YouTube lyric videos. Musicians want synced lyrics for Spotify releases. Karaoke users manually time tracks. AI music creators (Suno/Udio) need lyrics aligned after generation.

Existing solutions were expensive (often per-song pricing), unreliable (bad timing accuracy), overly complex desktop software, or locked to specific formats.

I wanted something simple: upload audio, get synced lyrics, download.

What I Built

LyricTime is an AI-powered lyric synchronization tool. Upload audio, transcribe lyrics, align timestamps to when each line is sung, export as LRC, SRT, or VTT.

Instead of an hour, it takes about 20-40 seconds.

The hard part wasn't transcription. Speech-to-text is easy now. The hard part was mapping each lyric line back to precise audio timestamps. I spent most of development time building a custom alignment system to get accuracy within ~0.1-0.3 seconds.

Users don't care how it works. They care that the line appears exactly when the singer starts singing.

The Business Model (And Why Two Versions Failed)

This part surprised me.

Version 1: Pay Before Use

Users had to purchase before trying anything. Conversion rate: ~1%. Nobody trusts a new tool enough to pay blind (except those 4 users that paid for a broken app).

Version 2: Free Minutes

Users got free processing minutes. Conversion rate: ~3%. The problem: most people only needed 1-2 songs. They used the free minutes and bounced. I was giving away what they needed.

Version 3: Free Preview (Current Model)

Users can upload, transcribe, edit, and see full results. But they must pay to download. Conversion rate: ~7-11%.

Why this works: they see full value, they invest time reviewing and editing, download becomes the natural final step.

The difference between "try everything free" and "preview everything free" changed the business.

What Actually Increased Conversion

Two things:

In-Browser Editing — Letting users tweak lyrics before downloading increased confidence, reduced support emails, made the product feel complete.

Removing Subscriptions — Switched to pay-as-you-go minute packs instead of recurring payments. Low friction. No recurring anxiety. No subscription guilt. $3 is an easy yes. So users just have have freedom to use the minutes whenever they want rather than having recurring payment.

What I Learned

Solve your own problem — I built this for myself first. That meant I knew what mattered and what didn't.

Launch before you feel ready — My first launch was accidental and unfinished. But real users gave real validation. Even broken validation is still validation.

Give value but not everything — If you give away the entire outcome for free, you remove the need to pay. Preview builds desire. Full access builds exit.

Listen to behavior, not assumptions — I assumed professional musicians were the target. Reality: TikTok creators, AI music users, karaoke hobbyists. The market tells you who your users are.

Current Status

180+ users. ~20 paid. Hundreds of thousands of seconds processed. Revenue growing month-over-month.

It's not life-changing money yet. But it's real. And it validated that the problem exists.

Final Thought

I didn't set out to build a SaaS. I just wanted to stop spending hours manually timing lyrics.

The biggest lesson wasn't technical. It was this: if a few people find and pay for something you barely launched, imagine what happens when you actually try.

If you're sitting on a half-finished project, ship it. You might forget about it. But someone else might not.

Oh and getting traffic from Google has been a pain in the ass. Bing brings 90% of my traffic. Hopefully when Google finally catches up it'll give me a decent bump in monthly visitors.

Forem: Marcus