A: Hey 👋

Thanks for the question.

So we bear the compute and storage cost of your training data as Tiny Talk and you as the owner of the API key, bear the cost of all API requests towards OpenAI.
Good news is, it's cheap.

Let's do a rough estimation:
General base unit for calculation is something called "token", a token generally corresponds to 4 characters of text.
The cost of one of the most expensive embedding model from OpenAI is $0.10 per 1 million token.

50,000,000 characters / 4 characters per token = 12,500,000 tokens.
(12,500,000 tokens / 1,000,000 tokens) * $0.10 = $1.25

So for "$1.25" you can train 50 Million character database

If we use the cheaper model, it is $0.02 per 1 million token.
Our cheapest cost for 50M Character DB would be $0.25

You see, it's really not even something to be concerned of.

Looking forward to have you on board with Tiny Talk!
Cheers 🌮