love romance 3D report

Couple chat, real shape.

Two years of anonymized synthetic couple chat, generated from daily events instead of polished love notes: errands, slow replies, short texts, quiet days, small fights, and repair attempts.

messages
52,3382024-01-01 to 2025-12-31
active days
729100% of the two-year span
avg per active day
71.8Includes quiet and conflict days.
longest silence
47.8hGap between two messages.
busiest day
2452025-08-06
sender split
A 60% / B 40%Close, but not behaviorally identical.

reality checks

The unglamorous parts are the signal.

This demo keeps sweet language restrained and gives weight to silence, logistics, media, short replies, and unfinished questions.

49%short or one-word messages
43%sent after 22:00 or before 02:00
301questions with no reply inside six hours
21days with ten messages or fewer
2,050images, voice notes, stickers, transfers, deletes
6%explicit affection-topic messages

frequency map

When the conversation lit up.

Click a day, hour, month, or word to open an aggregate insight tooltip.

Monthly volume

Daily texture

low
245

timing and balance

Late-night shape, starts, endings, and reply lag.

Volume imbalance is not affection. It mostly reflects work, bursts, silence, and who starts a thread after a gap.

Hourly rhythm

Peak hour: 23:00.

Monthly balance

real life topics

Mostly logistics, work, food, and waiting.

The synthetic generator now starts from daily states and events, so the topic mix is intentionally less romantic.

Topic mix

Message types

language fingerprints

The words that became habits.

Click a word to see who owns it and how strongly it shows up.

shared words

Contribution split, word by word.

Each row shows the same word split across both anonymized partners. Red leans Partner A, blue leans Partner B.

mood and affection

Soft signals, not verdicts.

Lexicon sentiment is rough. It reads words, not the relationship, and it should never decide who loves whom more.

Relationship keywords

evidence snippets

Tiny anonymized examples behind the charts.

Not a raw-message appendix: just a few synthetic excerpts that explain why the aggregate signals look the way they do.

Generated locally from 52,338 synthetic messages. This shareable report embeds aggregate metrics, not the raw message log. Treat real chat exports as private and anonymize before sharing.