How Edyt counts tokens

A few people have asked what "tokens" mean in Edyt's free plan.

It's a fair question, and slightly annoying that you have to ask it at all. You select some text, press a shortcut, and get a better version back. Nothing about that should require a unit of measurement. But the model underneath does have to count how much text it reads and writes, and that count is what your allowance is made of.

What a token is

A token is a small piece of text. Not quite a word, not quite a character.

In English it works out to roughly four characters, or about three-quarters of a word. Common words are usually one token on their own; longer or unusual ones get split into several. Spaces and punctuation count too.

You'll never need to work this out yourself. The only part worth remembering is that longer text costs more.

What gets counted

Two things, every time you run an action:

What you send. The text you selected, plus the instruction behind the action you picked.

What comes back. Edyt's response.

So it's just selected text + response. Fixing one sentence is cheap. Translating a long message or asking for a thorough explanation costs more, because there's more text moving in both directions.

What that looks like in practice

Rough numbers, not promises, but they'll give you the shape of it:

Most everyday edits land in the low hundreds.

How the free plan works

The free plan gives you an allowance that refreshes on a daily, weekly, and monthly cycle, so running out on a heavy afternoon doesn't cost you the rest of the week. The current numbers live on the pricing page rather than here, because they move as we tune the product and a blog post is a bad place to keep a number that changes.

Hit a limit and you're back in at the next reset. No credit card involved at any point. If you'd rather never think about it, Pro removes the cap.

Spending fewer of them

You really don't need to optimize this. But if you're a heavy user, three habits do most of the work:

• Select only what you're changing. One sentence costs a fraction of a whole document, and the result is usually better anyway, because the model isn't guessing which part you meant.
• Keep your custom prompts short. A tighter instruction tends to produce a tighter answer, which costs less coming back.
• Remember follow-ups are new requests. They're worth it when you need them, but three follow-ups is four actions.

That's the whole system. Edyt counts what goes out and what comes back, the free tier refills on a schedule, and Pro removes the ceiling.

If you want a feel for it, watch the counter after a few real edits. It'll make more sense in an afternoon of normal use than in any explanation of it, this one included.