go/no-go
kobi/thumblabmain · 2h ago
No ad account needed

What your code says you built

All of this came out of the repo. Fix anything wrong before it becomes ad copy — a bad read here wastes the whole test.

Read from source

23 files
Does what
Type a video title, get four YouTube thumbnails back in about ten seconds. Pick one, download it.
Evidence
app/api/generate/route.ts hits Replicate with an SDXL model · Generation model stores title, style, imageUrl
Sells how
Credit packs, one-off. Stripe has $9 / 25 credits, $19 / 100, $49 / 300. No subscription.
Tells
3 free credits on signup, so people can try before paying · a batch flag in the API that the UI never exposes · Clerk auth, so signup is two clicks
Sharpest edge
Ten seconds from title to four options. Everything else in this category makes you open an editor.
Costs you
About $0.04 in Replicate calls per generation. At $19 for 100 credits you keep roughly 79%.

Who might pay for it

Confirm before testing

Three guesses, ranked by how well your code fits them. Switch off any that are wrong.

Small YouTubers

Under 10k subs, making their own thumbnails in Canva at midnight. The obvious guess and the biggest crowd.

Fit

Faceless channel operators

Running 3–20 channels, publishing daily, need volume. That unused batch flag in your API is built for them.

Fit

Course creators

Coaches posting talking-head videos. Weakest guess — came from the category, not from your code.

Fit
Write the pages
go/no-go