do they like me?

reading screenshots…

loading the text reader…

it reads the text right here in your browser. the first time, it downloads a few mb of text-reading data.

Calculating interest...

reading the tone…

estimated time remaining: 3 situationships

how_it_works.txt - notepad

screenshots (easiest)

scroll to where you want to start then screenshot your way down to the most recent message. overlapping screenshots are fine and repeats get removed. drop them all in at once, in any order.

your messages are told apart from theirs by bubble color so this works for imessage, whatsapp, instagram and most other apps, in light or dark mode. screenshots only show date headers (like "Sat, Sep 19 at 1:25 AM"), not the time of each message, so the reply time readout is skipped. whatsapp and imessage exports have exact times, so reply times and late-night texts are real there. you get a chance to fix anything misread before it scores.

chat exports (most accurate)

drop the export anywhere on the desktop, or hit open chat export. it all gets read in this tab and nothing leaves it.

imessage (mac only)

apple has no export button so you need to use a free tool called imessage-exporter. it runs on your mac and nothing is uploaded.

open the terminal app (cmd + space, type terminal).
install the tool by pasting brew install imessage-exporter and hitting enter. if it says brew isn't found, download the mac version from the github link above instead.
give terminal permission to read your messages: system settings, privacy & security, full disk access, switch on terminal, then quit and reopen terminal.
paste imessage-exporter -f txt -o ~/imessage_export and hit enter. it makes a folder called imessage_export in your home folder with one .txt file per conversation.
open that folder (in terminal, open ~/imessage_export). find the file for the person. it's usually named after their phone number or email, like +15551234567.txt.
drag that one file onto this window or hit open chat export and pick it. don't drop the whole folder.

if it says the export path already contains export data, you've run it before. use a new folder name (like ~/imessage_export2) or delete the old folder first.

the folder has all your texts in it, so make sure to delete it when you're done.

whatsapp

one-on-one chats work best. you get a zip or txt, and you drop that file in as is.

iphone: open the chat, tap their name at the top, scroll down, tap export chat, choose without media. save it to files or airdrop it to your computer. you get a .zip.
android: open the chat, tap the three dots, more, export chat, choose without media. send the .txt to yourself (email works) and open it on your computer.
drag the .zip or .txt onto this window, or hit open chat export and pick it. no need to unzip it.

whatsapp stops at about 40,000 messages when media is off, so very long chats only include the most recent ones.

instagram

instagram makes you request your data first so start this ahead of time. it can take a few minutes or a few days.

on instagram, go to settings, accounts center, your information and permissions, download your information.
choose download or transfer information, pick your instagram account, then some of your information, and check only messages.
choose download to device, set the date range (all time is fine), set the format to json (not html), and submit.
wait for the email saying it's ready, then download the zip from the same page on your computer and unzip it.
open your_instagram_activity/messages/inbox/ (older exports: messages/inbox/). each chat is a folder named after the person plus some numbers.
open their folder and drag in message_1.json. if there's also a message_2.json, message_3.json and so on, select them all and drag them in together.

what it measures

every signal is scored 0 to 100 for both of you, then weighted into one index. a new conversation starts after 6 hours of silence.

tone (20%): every text gets sorted into love, happy, joking, frustrated, angry, cold or neutral using word lists and punctuation. warm texts add points, and cold, frustrated and angry ones take them away. one-word replies count as cold.
mirroring (18%): when they reply to you, do they match your text length, your lowercase habits and your emoji use.
starts the convo (14%): share of conversations they open. opening about half of them scores 70, and opening most of them scores 100.
asks questions (12%): real questions only. a question that's just logistics, like "where yall at," doesn't count. if 1 out of every 5 of their texts is a real question, that's a perfect score.
someday energy (14%): plans with an actual day or time score high. "we should", "next time" and "would love to" score low.
word effort (12%): share of all the words typed. an even split scores 70, and typing most of the words scores 100.
not left on delivered (10%): leaving your messages unanswered for a day or more.

the big number isn't a plain average of the whole chat. the whole chat counts for 40% of it, and the most recent part of the chat counts for 60%, because how someone texts lately matters more than how they texted months ago. the chart shows the same score over time, week by week (or day by day, or month by month, depending on how long the chat is).

extra (not used to calculate the score)

these are just for fun: the friend zone detector (only making plans in a group setting), mixed signals (how much their weekly score swings), median reply time, instagram posts, tweets and other links they send, and texts sent between midnight and 5 am.

the fine print

this is vibes math, not science. nobody has ever been loved or unloved by a number on a screen.