Slopify

Music thatnarrates your life.(in the worst possible way)

A peaceful trail becomes a metal rock concert.

We find the worst possible song for every occasion.

  1. See

    Camera glasses catch the scene

  2. Read

    It understands the moment

  3. Invert

    It finds the musical opposite

  4. Play

    Spotify plays the worst fit

A sunlit park

confidence 0.90

Bodies

Drowning Pool

Inverted, a pleasant park asks for funeral doom. Nu metal was the closest thing in the corpus — 0.911, the highest wrongness it scored.

A quiet library

arousal 0.08

Sandstorm

Darude

Nothing here is loud, so genre distance has nothing to work with. tempo_clash wins instead: a silent room cannot survive relentless arousal.

Pipeline playback

The scene becomes its musical opposite.

Footage from Meta glasses—or a video uploaded in the web app—enters the same pipeline. It reads the setting, maps its musical associations, and selects the genre and track that oppose the moment.

video input → opposite genre → track
Try it

Try it on a video.

Choose a sample or upload your own footage. It reads each moment and shows what it chose — and why.

Try it with your footageUpload your own videoOpen the demo workspace for analysis and synchronized playback.
The loop

One frame in.One wrong song out.

Every ~5 seconds. Two model calls at most — the rest is local.

The loop, as a diagramSix steps in sequence: Look (frame + 3s audio), Notice (local, no model), Understand (Gemini · 1.17s), Invert (reflect + taboo), Choose (judge · T 0.20), Commit (queue, or cut in). If nothing has changed, step two returns to step one without calling a model. Between Invert and Choose, six strategies run at once and a judge picks one. After Commit, the loop returns to Look, about every five seconds.every ~5 secondsnothing changedsix at oncegenre_antipodetempo_clashlyrical_ironysemantic_oppositeregister_clashcatalogue_dive01Lookframe + 3s audio02Noticelocal, no model03UnderstandGemini · 1.17s04Invertreflect + taboo05Choosejudge · T 0.2006Commitqueue, or cut in
Most frames never reach a model
The gate compares each frame locally. A still room loops without spending a call.
Six strategies, one judge
Genre, tempo, lyrics, meaning, occasion — and one that hunts outside the 47 songs.
Under 0.35 confidence it plays nothing
Pointed at a blocked camera it reported 0.10, and stayed quiet.
Results

What came out of it.

1.17s
Scene read, median
238
Tests
47
Songs, by hand
0
Images stored

  • It is never silent

    Every model, player and voice degrades to a stand-in, and under all of them sits a pre-picked fallback deck. Silence is the only real bug.

  • The reasoning is the product

    Watching it read a room correctly and then queue the worst possible answer is the whole thing in one screen. A shuffle button can't show its working.

  • Nothing here is faked

    Where something is a placeholder the interface says so. The bundled clips carry real Gemini sessions, and an upload either shows its own analysis or fails plainly.

Every decision, in full

Six moments, and what it did about them.

Read out of the recorded sessions in public/sessions/, in the order the clips play them — every read by gemini-3.5-flash-lite. Nothing is filtered or reordered: this is all of them, and six different songs out of a corpus of 47.

Reads it asConfidencePlaysBecause
university library study hall, group studying and taking notes at a long tableLibrary → Birthday → Gym · 00:00 · focused0.95DualitySlipknotquiet, academic, structured, indoor -> noisy, chaotic, impromptu, outdoor -> dance, metal, punksemantic_opposite · turned down Baby Shark, Crazy Frog
indoor domestic kitchen or dining room, birthday cake on a table, lighting candles on a birthday cakeLibrary → Birthday → Gym · 00:27 · joyful anticipation0.98O FortunaCarl Orffdomestic, informal, intimate, warm -> industrial, formal, public, cold -> ambient, classical, metalsemantic_opposite · turned down Ride of the Valkyries, In the Hall of the Mountain King
indoor gym or fitness studio training floor, performing pushups or plank exercise with athletic hand wrapsLibrary → Birthday → Gym · 00:39 · determined0.95In the Hall of the Mountain KingEdvard Griegutilitarian, industrial, high-energy -> ornate, residential, sedentary -> ambient, classicalsemantic_opposite · turned down Ride of the Valkyries, Clair de Lune
corporate office open-plan workspace, working at computer desks in an officeOffice → Food Court · 00:00 · focused0.95BodiesDrowning Poolarousal gap 0.91, meter clash 0.00tempo_clash · turned down Careless Whisper, Duality
shopping mall food court, people ordering food and dining in a busy food hallOffice → Food Court · 00:20 · bustling0.95Gymnopedie No. 1Erik Satiecommercial, casual, noisy, high-density -> secluded, formal, silent, sparse -> ambient, classical, operasemantic_opposite · turned down Clair de Lune, O Fortuna
snow-covered forest trail surrounded by dense pine trees, walking on a snowy path with an animalWinter Forest Walk · 00:00 · tranquil0.95Barbie GirlAquanatural, remote, quiet, wild -> artificial, urban, loud, domestic -> dance, disco, electronicsemantic_opposite · turned down Dancing Queen, September