Texts.Analyzer

Chat analyzer

Chat Analyzer: Message Counts and Reply Times

The generic version of the analyzer: any conversation, any app, pasted as text. It returns message counts, speaker share, reply times and the shape of the thread — and it draws a hard line at anything that would be a guess.

Stays in this tab. Nothing is uploaded or stored.

What it reports

The readout is deliberately plain. There are no badges, no colour-coded verdicts and no score out of ten. A thread that is 60/40 is not a better or worse thread than one that is 50/50; the number exists so you can see it, not so the page can judge it.

  • Total messages, number of speakers, and each speaker's share
  • Who opens a conversation and who answers first
  • Median reply time per person, with gaps over six hours ignored
  • Double texts, counted per person
  • Questions asked, per person
  • Sessions, the longest silence, and the longest single message

Why the generic mode matters

Most chat analyzers assume you have a WhatsApp export. Plenty of people do not: they have a copied iMessage thread, a Telegram block, a Messenger conversation they scrolled through, or a chat from an app nobody made an export for. The generic mode accepts all of them as long as each message starts with a name and a colon.

That also makes the tool usable for things that are not romantic at all — a freelancer checking whether a client thread drifted into one-sided replies, a group organiser seeing who actually responds to the plan, a researcher counting turns in a transcribed dialogue.

Reading the readout honestly

Start with volume, then timing, then sequencing. Volume tells you who is writing; timing tells you when they write; sequencing tells you who carries the conversation. Comparing all three over the same excerpt is what makes the numbers useful — a short reply sent in ninety seconds reads very differently from a short reply sent two days later.

Then stop. The data you have covers one excerpt of one channel. Whatever is happening in the relationship lives outside that excerpt, and no amount of counting changes it.

What the analyzer refuses to do

It will not produce an “interest level”, a compatibility score, or a prediction. Those require a model to assert meaning the text does not contain, and a confident-sounding number built that way is worse than no number. If you want a structured reflection instead, the quiz and the guides linked below are the honest versions of that.

Questions people ask before pasting a chat

Do I paste text or upload a file?

Both. Paste “Name: message” one per line, or open a WhatsApp .txt / .zip export. It also reads group chats and threads with more than two people.

How is reply time counted?

Reply time is the gap between someone else's message and yours, throwing away anything over six hours so a night's sleep does not inflate the average. The page shows the median, not the mean, because one slow week can drag a mean anywhere.

Is an uneven message count bad?

No. Long and short messages are both normal. The ratio is worth noting only when it changes: if the same two people wrote roughly equally for months and one side quietly dropped to a third, the change is the signal, not the ratio itself.

What happens to my chat afterwards?

Nothing. The text never leaves the tab, so there is no account tied to it and no history to look up.