voiceover.tn

A marketplace for voiceover artists in Tunisia. Real humans, not AI. That includes Derja, the Tunisian Arabic dialect that global voice tools still don't handle well.

I built it in public, streaming the work daily on YouTube. The site is live, matching works, AI-detection screening works. I paused it just before Ramadan 2026 to make a real call on what comes next.

The bet

AI voice generation is eating the global voiceover market. For English, French, MSA, it's fine. For Derja, the AI still sounds wrong to a native speaker. Brands shooting for a Tunisian audience were either using bad AI, casting through agencies (slow), or leaning on word-of-mouth (random).

The thesis was narrow on purpose: Derja, human only, verified by an AI-detection check on every uploaded voice. A marketplace small enough to win its niche fast.

Mechanics

• Compare 15 voices in 5 minutes. No per-creator messaging dance.
• AI detection on every voice. Only humans pass through.
• Free to post a job. 5% platform fee on the agreed price.
• Escrow. Payment held until the client approves delivery.
• Two free revisions baked into every brief

Built in public

I streamed the build daily on YouTube. Design decisions, stack choices, copy debates, all of it. People watching surfaced edge cases I would have missed in private. It also meant every shortcut I took had an audience, which made me cut fewer of them.

Why paused

Two things hit at once. Ramadan 2026 was starting (the Tunisian creative economy slows down for the month), and I needed to decide which side of the marketplace to fuel first. Funding both supply (artists) and demand (buyers) on a personal budget without a clear winner is how you burn runway in a hurry. Better to pause than half-fund both.

What's next

A short discovery sprint with one side of the marketplace (artists or buyers, not both) to figure out which lever moves things fastest. Then either restart, or wind it down honestly and write up the lesson. Either is fine.