I Launched a SaaS, Forgot About It, Then Discovered People Were Paying for a Broken Product
Source: Dev.to
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 kept getting emails that I assumed were just recurring charges from when I subscribed to the app myself.
Another month passed, and 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, announced it anywhere, or even remembered it existed.
When I opened the app, parts of it were broken, some features didn’t work, and the UI was pretty horrible. I was embarrassed, refunded everyone immediately, and sent apology emails with bonus transcription minutes (according to Convex, none of them have logged in and used the minutes).
But something stuck with me: if four people found my broken tool and paid for it, maybe there was something worth fixing. So I rebuilt it properly in 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 (e.g., 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. Users upload audio, the system transcribes the lyrics, aligns timestamps to when each line is sung, and exports the result as LRC, SRT, or VTT.
- What used to take an hour now takes about 20–40 seconds.
- Transcription itself is easy with modern speech‑to‑text models.
- The hard part was mapping each lyric line back to precise audio timestamps; I built a custom alignment system that achieves 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)
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 the four early users who paid for a broken app).
Version 2: Free Minutes
- Users received free processing minutes.
- Conversion rate: ~3%.
- Most people only needed 1–2 songs, 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 must pay to download.
- Conversion rate: ~7–11%.
Why this works: Users see the full value, invest time reviewing and editing, and downloading becomes the natural final step. The difference between “try everything free” and “preview everything free” changed the business.
What Actually Increased Conversion
- In‑Browser Editing – Letting users tweak lyrics before downloading increased confidence, reduced support emails, and made the product feel complete.
- Removing Subscriptions – Switching to pay‑as‑you‑go minute packs instead of recurring payments lowered friction. A $3 purchase is an easy “yes,” and users have the freedom to use minutes whenever they want without recurring anxiety.
What I Learned
- Solve your own problem – Building this for myself first meant I knew what mattered and what didn’t.
- Launch before you feel ready – My accidental, unfinished launch gave real validation. Even broken validation is still validation.
- Give value but not everything – Giving away the entire outcome for free removes the incentive to pay. A preview builds desire; full access drives purchase.
- 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 is growing month‑over‑month
It’s not life‑changing money yet, but it’s real and validates 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.
Note: Traffic from Google has been a pain; Bing currently brings about 90 % of my visitors. Hopefully Google catches up and gives me a decent bump in monthly visitors.