
Let’s make time travel real! For my latest project I’ve been working a lot (as in… a LOT) with Google Veo — and I stumbled upon something fascinating: When I asked characters to speak a language that no longer exists… they actually said something. I’m a massive history nerd so I had to run a little experiment.
Disclaimer: Don’t take it too seriously and obviously I know AI output like this is riddled with mistakes. This is for fun and people who love languages and history!
🧪 The Experiment
I started by selecting a set of extinct or rarely spoken languages. Then I generated historically inspired character portraits — fixing or photoshopping some obvious inaccuracies (AI loves old Hollywood clichés). Finally, I fed Veo a simple video prompt template:
static camera shot of the person saying directly into the camera in [language]: “Greetings, I am [typical culturally appropriate name], and this is my attempt at speaking [language].”, then the person smiles, no music
Here’s the LATIN version
Prompt:
static camera shot of the person saying directly into the camera in Latin: “Greetings, I am Marcus Licinius Macer, and this is my attempt at speaking Latin.”, then the person smiles, no music
Veo’s output:
Salvete, ego sum Marcus Licinius Macer, et hoc est conamen meum Latine loquendi.
❌ Issues
- Pronunciation mistakes (e.g., “hoc” sounds like “hos”)
- Ecclesial/Italian accent (still better than English accent!)
- Roman naming conventions originally ignored by LLM (manual adjustment)
✅ Things that actually worked
- Syntax surprisingly close to correct
- Tone felt natural — even if imperfect
💬 I’d love your input
Do you know someone who might enjoy analyzing this further?
And what interesting use cases could this unlock?
Which other languages do you want to hear?
Did you know AI speaks ancient languages?
byu/Storybook_Tobi inChatGPT
