Missed Me by Thaaaat Much!

I was approached by a company last week to record 50,000 words for their text-to-speech (TTS) engine. The stated use? eLearning for kids and young adults. The money wasn’t terrible, especially since I’ve had a down year in the eLearning voice over genre, but a few things immediately made me pause.

First, the company name wasn’t visible or offered. Now, I had recorded a TTS engine years ago for my previous employer, before I went full time into VO. The project never fully launched, but I learned a lot from that experience. I thought I could do a solid job, but I also know that TTS is basically the grandfather of AI, so I approached this cautiously.

Sidebar: If you’re not familiar, TTS works by joining phonemes to create words. Phonemes are the smallest building blocks of speech. They're the tiny sounds that make up words. When you say “cat,” you’re actually saying three phonemes: /k/ + /æ/ + /t/. Text-to-speech systems record and analyze thousands of these tiny units so they can rebuild words, sentences, and even mimic emotions later. It's kind of like assembling a puzzle from very small pieces of your voice.

So, I asked for more information. I wanted the company name, the hosting platform (because I very much doubted they had their own TTS engine, That's why my old employer's project never got off the ground). I also wanted a chance to review the agreements. And that’s when the red flags went up.

The hiring company was Chinese. Googling their name yielded nothing exact, just vague results. The equivalent of searching for McDonald's and finding, “Unilever presents McDonald’s” or “Adobe presents McDonald’s.” Confusing at best. And the delivery timeframe? “Due in 90 (sixty) days.” Huh? The hosting company, however, was crystal clear… and scary.

Reviewing the contracts didn’t help. My spidey sense was tingling. So, I uploaded them to ChatGPT to help untangle the legal jargon. Sure enough, my instincts were right.

Here’s a snapshot of the clauses that made me pause:

• Clause (6) – Total transfer of rights You’d give the hosting company all rights to your recordings “for commercial and any other use,” including AI voice cloning or speech synthesis. You may not use your own recordings. Essentially, this is a full voice-model buyout—globally and forever.

• Clause (13.1) – IP ownership

  They claim all intellectual property rights, including patents and copyrights, to everything you record. This gives them legal ground to use your voice indefinitely.

• Clause (15) – Indefinite validity The contract technically expires in 2026, but the rights you transfer continue forever.

• Clause (12B4) – Punitive late fees A 3% per-day penalty for late delivery could wipe out the entire contract value in just days—unreasonable and far from standard for freelance voice actors. And how are we supposed to know when it's late 60 days? 90 days?

• Payment timing (Clause 11) You deliver 20% of the project before receiving 40% of payment, and final payment could be delayed up to 90 days after acceptance. There’s no guarantee the files will ever be “accepted”—they could request endless retakes. And seriously, most of the language implied, "we'll find a way to reject your audio so you'll be caught in an endless cycle of recording and never being paid.

• Clause (10) – Mandatory future work You’d be bound to accept additional recording tasks for a year at the same terms, with no option to renegotiate or decline.

• Jurisdiction & enforceability The governing law is Chinese. For a U.S.-based freelancer, this means no practical recourse if things go wrong.
  

And that’s just the first document. The second contract was even worse:

• Clause 1 – Sale of your voice You’re authorizing them to transmit, share, and sell your recordings to third parties for AI research, algorithm training, and commercial purposes, forever.

• Clause 2 – Global, perpetual authorization They can use and license your voice worldwide, indefinitely.

• Clause 4 – Data storage and resale Your recordings would be stored, repackaged, and sold, completely out of your control.

• Clause 7 – Legal jurisdiction Any legal dispute would require suing in China, a near impossibility for most freelancers.
  

The “data protection” sections read like window dressing. Encryption and data officers are mentioned, but the contracts explicitly allow transmitting your voice to third parties.

I politely declined and wished them luck. I felt relieved. I defended my voice. Yay!

Until the next day, when they came back saying they could limit it to eLearning modules and increase the budget by 50%. But even then, my long-form narrations could be used for TTS or AI voice cloning. Excuse me, is there a “kick me” sign on my back?

I declined again. It stings a little, because accepting this one contract could have made up for a third of the eLearning work I’ve lost this year to AI. But standing up for yourself, and protecting your voice, rights, and long-term freelance career, well, it's priceless.

For any freelance voice actor navigating AI and TTS offers, short-term gains might look tempting, but know that giving away your rights could cost far more in the long run. They're counting on ignorance. Don't give them the satisfaction.