50 AI Playlist Prompts That Actually Work in 2026 (Tested on Spotify)
50 evocative, tested prompts you can paste into an AI playlist generator to get back a playlist that doesn't feel algorithmic — covering moods, decades, niches, micro-scenes, and weird specifics.
By Gabin Fay
50 AI Playlist Prompts That Actually Work in 2026 (Tested on Spotify)
Why Your Playlist Prompt Is Probably Underperforming
Most people approach an AI playlist generator the same way they approach a search bar: they type "chill vibes" or "workout music 2026" and hit enter. The result is technically a playlist. It has songs. The songs play. But it feels assembled, not curated — the musical equivalent of an airport lounge. No thread. No intention. Just 25 tracks that all kind of fit a vague category.
Here is the uncomfortable truth: the playlist is only as interesting as the prompt that generated it.
An AI playlist generator is not Spotify's algorithm. It does not optimize for engagement metrics or keep you in a safe, familiar bubble. It reasons about music — about sonic texture, cultural context, historical moment, emotional register. Feed it a rich prompt and it will return something that feels like it was made by a person who cares about music. Feed it "chill vibes" and it will give you a technically correct answer to a question that was never interesting in the first place.
So what separates a weak prompt from a strong one? Three things.
Specificity of context. "Lo-fi hip-hop for studying" describes a genre and an activity. "Lo-fi hip-hop that sounds like it was made in a Tokyo apartment at 2am while it's raining outside, recorded on cassette" describes a world. The second version gives an LLM enough to triangulate: geography, time of day, weather, texture, recording medium. Each detail eliminates thousands of wrong answers and points toward the right ones.
Emotional precision over emotional category. "Sad music" is a category. "The specific melancholy of watching someone you love drive away knowing you won't see them for a long time" is an emotion. One of those is solvable. The other is boring. Music that captures a specific emotional nuance tends to come from a much more interesting corner of the catalog than music that captures a broad emotional category.
Historical and cultural anchoring. Mentioning a year, a city, a scene, or a subculture transforms a prompt from abstract to concrete. "Post-punk" could mean 500 different things. "Post-punk from Manchester between 1978 and 1983, before the synths took over" means almost nothing except exactly the right thing. The AI has enough context to build a playlist that feels like it was assembled by a music journalist who covered that scene.
The 50 prompts below are organized into seven sections, moving from place-and-time anchors through mood and activity, into era and scene specifics, edge cases and constraints, niche genres, cinematic vibes, and finally the deeply personal. Each prompt is written the way you would actually type it into Playgen — not like a Google search query, but like a description you might give to a record store clerk you trust. Each one includes a note on why it works and, where relevant, a few example artists or tracks you might expect to see in the results.
These are not templates. They are invitations. Modify them, collapse two of them together, replace the city with your city. The goal is not to use these prompts verbatim — it is to understand how a good prompt thinks, so you can write your own.
Section 1: The Place & Time Prompts
These prompts work because geography and time are music's most reliable context-setters. A city in a specific decade carries a sound with it almost automatically: the labels that operated there, the clubs that shaped the scene, the economic conditions that pushed artists in particular directions. A strong place-and-time prompt gives an AI more to work with than almost any mood description.
Prompt 1: Late August in a Lisbon kitchen, 2:30pm
A Sunday afternoon in a Lisbon apartment — the shutters half-closed, a pan of something on the stove, fado drifting from an old radio mixed with the kind of Brazilian MPB that sounds like it was recorded outside. Unhurried. Slightly melancholic. No urgency.
Why it works: Lisbon and Brazil share a musical and linguistic lineage that generates a very specific sound — melancholic but warm, acoustically intimate, rhythmically relaxed without being passive. The kitchen and the afternoon give a setting that filters out anything too produced or too urgent.
Expect: Mariza, Gilberto Gil circa 1978, Caetano Veloso's quieter moments, Ana Moura.
Prompt 2: A 1997 dial-up modem ritual
The sound of connecting to the internet in 1997 — alt-rock and post-grunge with a low-energy, slightly anxious undertone. Songs that were on the radio while you waited for a webpage to load. Nothing that requires full attention. A kind of ambient mid-90s American malaise.
Why it works: The specificity of the year and the activity pulls the AI away from generic 90s nostalgia and toward a very particular sonic atmosphere — the tail end of grunge, the beginning of something softer. The word "malaise" does a lot of tonal work.
Expect: Matchbox Twenty, early Third Eye Blind, Foo Fighters B-sides, Semisonic.
Prompt 3: A Glasgow pub on a wet Tuesday in November
The kind of Scottish pub where nobody is performing happiness — conversation is low, the beer is real, and the jukebox is playing something that sits between folk and indie rock. Songs with weight but not drama. The kind of music that makes silences between people feel comfortable rather than awkward.
Why it works: The emotional register is unusually specific — not sad, not celebratory, but somewhere in between. Scottish folk-rock occupies that space almost by definition, and the "wet Tuesday in November" sets a mood that filters out anything too energetic.
Expect: Frightened Rabbit, The Proclaimers, Teenage Fanclub, Arab Strap.
Prompt 4: Tokyo at 4am after the last train
Walking through Shibuya or Shinjuku when the crowds are gone and the neon is still on. The city is quiet but not empty. Something electronic that feels urban and slightly lonely — not club music, not ambient, but something in between. Japanese or non-Japanese, but it has to feel like that specific hour.
Why it works: The after-hours, post-crowd urban setting is a powerful filter. It points toward a particular texture of electronic music — produced for late nights but not for dancing, with an urban loneliness that Japanese electronic artists in particular have explored.
Expect: Cornelius, Nujabes, Burial (early), Telefon Tel Aviv.
Prompt 5: A summer afternoon in 1980s New Orleans
The heat of a New Orleans afternoon in the mid-1980s — second-line rhythms, early zydeco, some blues that feels properly humid, and the kind of funk that came out of the Crescent City before the rest of the country caught up. Nothing polished. Nothing from a studio that was trying to go national.
Why it works: New Orleans music has a specific regional identity that national-facing production tended to sand down. This prompt explicitly asks for the local, unpolished version, which is an instruction an AI can act on.
Expect: The Meters, Clifton Chenier, Professor Longhair, early Trombone Shorty predecessors.
Prompt 6: Arriving in Berlin in 1992 with a one-way ticket
The music you would have heard in Berlin in 1992 if you arrived from somewhere in Eastern Europe with nothing. Industrial techno, the first wave of trance before it got saccharine, some post-Wall ambiguity — sounds that feel like a city still deciding what it wants to be. Not polished. Not victorious. Uncertain.
Why it works: This prompt does something unusual — it assigns emotional stakes to a historical setting. The "one-way ticket" and "Eastern Europe" locate a specific experience within the broader Berlin music story.
Expect: Tresor-era techno, early R&S Records output, Basic Channel, Klaus Schulze late period.
Prompt 7: A Buenos Aires milonga at midnight
Traditional tango played at a milonga at midnight — not the tourist show version, not the neotango-electronic fusion, but the real thing. Orchestras from the Golden Age. The kind of music where the dancers are listening as hard as they are moving. Dignified and devastatingly romantic.
Why it works: Specifying "not the tourist show version" is a useful instruction to an AI — it rules out an entire category of results that would otherwise dominate. The Golden Age constraint adds historical precision.
Expect: Astor Piazzolla early period, Aníbal Troilo, Juan D'Arienzo, Carlos di Sarli.
Prompt 8: Nashville, 1974, before the crossover
Country music from Nashville around 1974 — before the Urban Cowboy era sanitized it, before the crossover acts started softening the edges. Classic country with real string arrangements, lyrics about trucks and whiskey and actual heartbreak, produced with taste but not slickness.
Why it works: The "before the crossover" instruction is doing most of the work — it tells the AI to avoid a specific later development and stay in an earlier, rawer moment.
Expect: Merle Haggard, Loretta Lynn, George Jones, early Waylon Jennings.
Section 2: The Mood & Activity Prompts
Activity prompts are the most commonly searched category ("music for working out," "music for studying") and usually the most generic. The trick is to describe the specific version of the activity — not "studying" but the particular texture of a long, late-night session with a specific kind of focus.
Prompt 9: Folding laundry on a slow Sunday
Coffee-shop indie folk for the most undemanding domestic task. Songs that feel like a half-read novel — present but not demanding attention. Acoustic, slightly warm-recorded, with the kind of vocals that are pleasant but not show-offy. Nothing with a big chorus that will interrupt the rhythm of folding.
Why it works: The explicit "nothing with a big chorus" is a negative constraint that an AI can act on. The domestic, unhurried specificity rules out anything with urgency.
Expect: Iron & Wine, early Sufjan Stevens, Bon Iver's quieter moments, Gregory Alan Isakov.
Prompt 10: An hour-long deep-work session that never breaks tension
Minimal techno at exactly the kind of tempo where you stop noticing it's playing — somewhere between 114 and 118 BPM, without breakdowns that interrupt focus, without vocals, without melodies that are interesting enough to pull you out of the work. Relentless but not aggressive. A machine that runs in the background of your brain.
Why it works: The BPM specification and the "no breakdowns" instruction give an AI very precise parameters. The final metaphor — "a machine that runs in the background of your brain" — sets a tonal target that is harder to quantify but equally useful.
Expect: Surgeon, Plastikman, early Regis, Drumcell.
Prompt 11: Running uphill in the rain
Music that matches the specific unpleasantness and satisfaction of running uphill in the rain — not pump-up anthems, but something that feels like controlled suffering. Mid-tempo rock with real drums, maybe some post-punk urgency, nothing that asks you to feel good about what you're doing. Music for grinding, not for triumph.
Why it works: "Not pump-up anthems" is a crucial negative constraint — it rules out the most common result category for running music. "Controlled suffering" is an unusual but precise emotional target.
Expect: Refused, Protomartyr, Drive Like Jehu, Idles.
Prompt 12: Cooking something that takes three hours
Background music for a long, slow cook — braising something, probably. Something that evolves over time rather than cycling through the same energy. Jazz that starts relaxed and gets more complex, or soul that shifts from afternoon warmth to evening depth. The playlist should feel like it tracks the hours.
Why it works: The "evolves over time" instruction asks for something architecturally unusual — a playlist with an arc, not a static mood. That is something an AI can plan for in a way a simple shuffle cannot.
Expect: Bill Evans to McCoy Tyner trajectory, Nina Simone, classic Marvin Gaye B-sides.
Prompt 13: The walk between two important conversations
The ten minutes between leaving one significant conversation and arriving at another — the transitional headspace music. Something that neither sustains the previous mood nor prepares you for the next one. Ambient or near-ambient, slightly disconnecting, no strong emotional color. A sonic palate cleanser.
Why it works: This is a genuinely unusual use case that requires a genuinely unusual sonic texture. "No strong emotional color" is a precise instruction for ambient music selection.
Expect: Brian Eno's ambient series, Stars of the Lid, Tim Hecker's quieter work.
Prompt 14: Reading a thriller on a long flight
Music without words, with enough tension to sustain concentration on a plot-heavy book but without enough drama to distract from it. Film scores that are designed to accompany action but not draw attention to themselves. The kind of music that speeds time without hijacking it.
Why it works: "Enough tension to sustain concentration but without enough drama to distract" is a precise calibration that eliminates both the boring and the over-dramatic.
Expect: Jonny Greenwood film scores, Cliff Martinez, Ennio Morricone's quieter work.
Prompt 15: The hour after a breakup when you're still in shock
Not the crying-in-bed phase. The first hour, when you're still numb and the body is running on autopilot. Music that matches the shock without accelerating the grief — not sad, not angry, but slightly dissociated. The kind of music that plays when you're not quite inside yourself yet.
Why it works: This is a specific emotional phase that rarely gets named, let alone scored. The "not sad, not angry" negative space is actually very precise and points toward a particular dissociative sonic register.
Expect: The XX, Bon Iver's "For Emma," Beach House's darker moments.
Prompt 16: A long drive where no one speaks
Music for a car full of people who are not talking — not because there's tension, but because everyone has agreed, silently, that the music is sufficient. Something that gives the landscape meaning without demanding attention. Road-trip music for introverts. No singalongs, no anthems.
Why it works: The social context (a group choosing silence) is unusual and creates a specific tonal requirement — music that can hold a room without performing.
Expect: Mazzy Star, Nick Drake, Gillian Welch, Sparklehorse.
Section 3: The Era & Scene Prompts
Scene-specific prompts are where AI playlist generators show their musical depth. A well-specified historical scene has enough documented information — labels, clubs, key releases, associated artists — that an AI can build something that feels like genuine curation rather than keyword-matching.
Prompt 17: Pre-internet British rave — 1989 to 1992
The acid house and early hardcore rave scene in Britain before it went commercial — the Haçienda, the M25 orbital raves, the pirate radio stations, the moment before the sound split into jungle and trance. Roland 303 basslines, breakbeats that haven't been programmed yet, the rawness of a scene that didn't know it was going to become the foundation of everything.
Why it works: The end date (1992) and the reference to what came after ("before the split") give the AI a historical boundary that rules out the more polished later-era output. The pirate radio and orbital rave references add cultural texture.
Expect: The Prodigy's earliest EPs, LFO, 808 State, Altern-8.
Prompt 18: Detroit techno's second wave, 1992–1995
Not Juan Atkins and the Belleville Three — the second generation. The moment before the Europeans arrived and turned it into something more minimal and aesthetic. Underground Resistance's more aggressive period, the Model 500 follow-ups, the tracks that were being made in warehouses and never meant to be album tracks.
Why it works: Specifying "not the founders" and naming the temporal window before external influence transformed the scene is a sophisticated historical instruction that rules out the most commonly cited material.
Expect: Underground Resistance, Drexciya, Carl Craig's early work, Kenny Larkin.
Prompt 19: The Rough Trade moment in London, 1979–1983
The independent label ecosystem around Rough Trade in early-80s London — post-punk that was consciously political but also melodic, the intersection of the Gang of Four ethic with something more tuneful. Bands that were reading Marxist theory and also writing three-minute songs. The sound of a DIY infrastructure being built.
Why it works: Naming a specific label and its cultural context gives an AI a very well-documented corner of music history to work in. The "reading Marxist theory and also writing three-minute songs" is a tonal description that's both funny and precise.
Expect: The Raincoats, Essential Logic, Young Marble Giants, The Slits, Scritti Politti early period.
Prompt 20: São Paulo's tropicália underground, 1968–1972
The Brazilian tropicália movement at its most experimental — the moment before Caetano Veloso and Gilberto Gil were exiled and the government cracked down. Electric guitars meeting baião rhythms, concrete poetry as lyrics, music that was deliberately too strange for its political moment. Not the famous records — the margins of the movement.
Why it works: "Not the famous records — the margins of the movement" is an instruction that specifically avoids the canonical and asks for the peripheral, which an AI with broad music knowledge is well-placed to find.
Expect: Tom Zé, Gal Costa's stranger recordings, Os Mutantes B-sides, Capinan collaborations.
Prompt 21: Seattle in 1993 — not the famous bands
The Seattle sound in 1993, one year after Nirvana's moment, but specifically not Nirvana, Pearl Jam, or Soundgarden — the bands that were operating in their shadow. The second tier of the scene: harder, weirder, less concerned with being palatable. The bands that didn't get the major label calls but were arguably more interesting.
Why it works: The explicit exclusion of the canonical bands is both a creative constraint and a musical instruction — it pushes the AI toward the periphery of a well-documented scene, which is often where the most interesting music lives.
Expect: Mudhoney, Green River recordings, Skin Yard, Gas Huffer.
Prompt 22: Post-Soviet Tbilisi electronic scene, 1994–2000
The emergence of electronic music in Tbilisi after the Soviet collapse — recorded with whatever equipment was available, distributed on cassettes, influenced by what was filtering in from Western Europe but filtered through a very different cultural and economic reality. Rough, improvised, genuinely underground in the literal sense.
Why it works: This is a genuinely obscure historical moment that rewards an AI's breadth of knowledge. The "whatever equipment was available" and "cassette distribution" instructions point toward a specific sonic texture of constraint.
Prompt 23: New York no-wave, 1977–1981
The deliberately difficult, anti-commercial aftermath of punk in lower Manhattan — music that was sometimes noise, sometimes art installation, sometimes performance, and occasionally all three simultaneously. The Contortions, DNA, Glenn Branca — music made by people who had decided that the guitar was a percussion instrument and melody was optional.
Why it works: No-wave is small enough as a historical moment that precise prompting is less important than knowledge — the key value is naming the aesthetic principles ("the guitar is a percussion instrument") rather than the artists.
Expect: The Contortions, Lydia Lunch, Mars, DNA.
Section 4: The Edge-Case & Constraint Prompts
These prompts work by giving an AI an unusual technical or structural parameter to work within. The constraint itself becomes the curation mechanism — instead of describing a feeling, you describe a rule, and the rule does the filtering.
Prompt 24: Songs in 7/8 time that don't feel like prog
Tracks in 7/8 time signature that manage to not sound like a progressive rock exercise — music where the odd meter is structural rather than performative, where you might not notice it's in 7/8 until you try to count along. Folk, electronic, or post-rock that uses the time signature without announcing it.
Why it works: The "without announcing it" instruction is the key — it distinguishes between music that uses an unusual time signature as a feature and music that uses it as a foundation. The former is showing off; the latter is interesting.
Expect: Sufjan Stevens ("Come On! Feel the Illinoise!"), Radiohead ("Pyramid Song"), some Aphex Twin work.
Prompt 25: Tracks under 90 seconds that contain a full emotional arc
Songs that are shorter than a minute and a half but feel complete — not sketches, not interludes, but fully realized pieces of music with a beginning, development, and resolution. Hardcore, classical miniatures, ambient fragments that somehow feel whole. The challenge is the density.
Why it works: Duration as a constraint is inherently interesting because it forces the AI to consider music across very different genres that share a structural property. It also surfaces genuinely unusual choices.
Expect: Some Black Flag tracks, Bartók miniatures, certain Aphex Twin ambient works, some Slint moments.
Prompt 26: Songs recorded in one take that you can tell
Music where you can hear the room — where the imperfections are part of the texture, where a breath before a vocal line or a slight timing variation between musicians is evidence of a real moment. Not lo-fi affectation, but actual live-to-tape humanity. The anti-Pro Tools playlist.
Why it works: "Where you can hear the room" is a sonic description rather than a genre description — it cuts across rock, jazz, folk, and country to find a specific recording philosophy.
Expect: Bill Withers live recordings, certain John Prine sessions, early Townes Van Zandt, some PJ Harvey recordings.
Prompt 27: Electronic music that uses silence as an instrument
Tracks where the gaps between sounds are as compositionally important as the sounds themselves — electronic music that breathes, that lets you hear the negative space. Not ambient drone, but music where a silence of two beats is a compositional choice with consequences. Rhythm constructed from absence.
Why it works: This is a specific compositional principle that an AI with knowledge of electronic music production can apply across genres. The "rhythm constructed from absence" formulation is precise.
Expect: Burial, some Autechre work, early Massive Attack production, King Krule's more experimental moments.
Prompt 28: Covers that are better than the original and everyone knows it
Not debatable cases — genuine consensus that the cover outperformed the source material. The version that erased the original from cultural memory. Songs where the songwriter was outplayed by an interpreter who understood something the writer didn't.
Why it works: This is an unusual framing that points toward a very specific and well-discussed subset of recordings. The "everyone knows it" qualifier further filters toward consensus cases rather than controversial opinions.
Expect: Jeff Buckley's "Hallelujah," Jimi Hendrix's "All Along the Watchtower," Johnny Cash's "Hurt."
Prompt 29: Songs whose intros are longer than most people's attention spans
Tracks where the actual song doesn't start for two minutes or more — long intros that are interesting enough to reward patience. Music that refuses to announce itself quickly. The opposite of the streaming-era hook-in-five-seconds philosophy.
Why it works: This constraint explicitly inverts a dominant cultural norm (short intros for streaming engagement) and points toward an older recording philosophy or deliberately contrarian modern work.
Expect: Led Zeppelin's "Babe I'm Gonna Leave You," some Godspeed You! Black Emperor tracks, Arcade Fire's "Funeral" suite openers.
Prompt 30: Songs that would soundtrack a film's most ambiguous scene
Music for the scene where you genuinely don't know if something good or terrible just happened — the outcome is unclear, the emotional register is suspended. Not sad, not triumphant, but hovering. The kind of music that plays while the camera holds on a face that isn't moving.
Why it works: This prompt requires music with genuine tonal ambiguity, which is a specific compositional quality that filters aggressively. Most music resolves emotionally — this prompt asks for music that doesn't.
Expect: Jonny Greenwood's more dissonant film cues, some Ennio Morricone, Arvo Pärt, late period Harold Budd.
Section 5: The Niche-Genre Prompts
Genre prompts fail when the genre is too broad. They succeed when the genre is narrow enough to have a specific sound, a defined moment, and a set of artists that most algorithms don't know how to surface. These prompts are for the corners.
Prompt 31: Vaporwave that takes itself completely seriously
Vaporwave that is not ironic — music in the genre that treats its nostalgia and its sonic sources as genuinely moving rather than as aesthetic material for detached commentary. Not parody, not camp, but actual emotional investment in a degraded 1980s office aesthetic. The minority report of vaporwave.
Why it works: Most vaporwave operates with ironic distance; this prompt asks for the subset that doesn't. That is a specific and defensible aesthetic position.
Expect: Macintosh Plus's more earnest moments, some Saint Pepsi work, Nmesh.
Prompt 32: Estonian Soviet-era underground rock, 1986–1990
Rock music recorded in Soviet Estonia in the late 1980s as the independence movement was building — music that was underground not by genre definition but by political necessity. The sound of a culture asserting itself through electric guitars and lyrics that the censors couldn't quite object to.
Why it works: This is a specific enough historical and geographical moment that an AI with broad music knowledge can locate actual recordings, even if they are obscure. The political framing adds cultural texture.
Expect: Ruja, Ultima Thule, In Spe — bands that operated in the pre-Glasnost thaw.
Prompt 33: Jazz-funk from the Black-owned LA labels, 1972–1978
The jazz-funk and soul-jazz coming out of Los Angeles in the mid-1970s, specifically from the independent Black-owned labels that were operating before the majors noticed the genre existed. Rougher than what got released nationally, more rhythmically aggressive, less produced.
Why it works: The label ownership specification is a genuine music history distinction that points toward a different kind of sound than the better-known national releases. The "before the majors noticed" temporal boundary is precise.
Expect: Bobby Hutcherson's Blue Note output, early Crusaders recordings, some Minnie Riperton sessions.
Prompt 34: Isolationist ambient — the stuff that was too dark for Eno
Ambient music from the early 1990s scene that took Brian Eno's principles and pushed them toward something more threatening — music intended for listening in a dark room alone, not for comfortable background listening. The genre that the ambient journals called "isolationism." Texturally difficult. Emotionally uncomfortable.
Why it works: Naming the specific critical term "isolationism" anchors the prompt in a documented historical debate and gives the AI a genre label to work with.
Expect: Robert Hampson's work, Kevin Drumm, some early Coil ambient recordings, Main.
Prompt 35: Appalachian folk recorded before the field researchers arrived
The earliest available recordings of Appalachian folk music — before the Library of Congress expeditions normalized the repertoire, before the folk revival of the 1960s polished the edges. The music as it was actually practiced: irregular, unaccompanied sometimes, with regional variations that got averaged out later.
Why it works: The "before the field researchers arrived" instruction asks specifically for the pre-documentation era, which pushes toward genuinely archival material and away from the revival-era canonization.
Expect: Library of Congress recordings, some Lomax field recordings, the Carter Family's earliest output.
Prompt 36: Cumbia villera — the Buenos Aires underground, 2000–2006
Cumbia villera from Buenos Aires in the early 2000s — the working-class variant of cumbia that emerged from the villas miseria, lyrically raw, rhythmically insistent, treated as embarrassing by the mainstream Argentine media and beloved by everyone else. The music of a city in economic collapse.
Why it works: The social context (villas miseria, economic crisis) adds a layer of meaning that makes the musical output legible. The "treated as embarrassing by mainstream media" framing captures a specific cultural dynamic.
Expect: Damas Gratis, Pibes Chorros, Yerba Brava.
Prompt 37: British psychedelic folk — 1968 to 1973, the acid-pastoral moment
The brief, strange period of British psychedelic folk — not rock, not trad folk, but the intersection: acoustic instruments, lysergic production, lyrics drawn from mythology and the English countryside, a profound seriousness about both folk traditions and experimental music. Pentangle, Fairport Convention, and everything around them.
Why it works: "Acid-pastoral" is a specific critical term for this moment, and naming it gives the AI a precisely located genre with a documented canon and periphery.
Expect: Pentangle, Fairport Convention, Anne Briggs, Nick Drake's early period, Trees.
Section 6: The Cinematic & Vibe Prompts
These prompts work not by naming a historical moment but by evoking a scene that doesn't exist — a film that was never made, a sequel to a mood, a specific imaginary setting. The AI's job is to find the music that would score that unrealized scene.
Prompt 38: Credits music for a French New Wave film that was never made
What would play over the closing credits of a Godard or Varda film from 1963 that doesn't exist — a film about two people who almost fell in love in Paris, with an inconclusive ending. Jazz that feels specifically French in its restraint, possibly with a theme that was heard earlier in the film in a different key.
Why it works: The fictional film conceit gives permission for a very specific tonal target without requiring real examples. "A theme heard earlier in a different key" asks for music with a sense of variation and history.
Expect: Miles Davis's "Ascenseur pour l'échafaud" recordings, Michel Legrand, some Martial Solal work.
Prompt 39: The library music section that nobody touched
Library music from the 1970s — the recordings made for TV and film that were filed and rarely used, sitting in catalogs that nobody knew about until the internet made everything searchable. Orchestral, electronic, experimental arrangements made by composers who were solving a brief rather than making art, and accidentally made art.
Why it works: Library music as a category is well-enough documented that an AI can locate specific catalogs and composers. The "nobody touched" instruction points toward the more unusual corners.
Expect: KPM catalog recordings, Syd Dale, some Piero Umiliani work, certain Alessandro Alessandroni pieces.
Prompt 40: A party that started well and is now winding down
The last 45 minutes of a good party — not sad, just deflating. The energy is lower, the conversations are more honest, a few people are already checking when the trains run. Music that acknowledges the evening is ending without rushing it. Not outro music. The dying ember phase.
Why it works: "The dying ember phase" is a precise emotional and temporal location. The "a few people are already checking when the trains run" image is specific enough to orient the AI's tonal choices.
Expect: LCD Soundsystem's slower moments, some Arthur Russell, Yo La Tengo late-night recordings.
Prompt 41: What winter sounds like in a country that doesn't understand winter
Music that captures the experience of winter in a place that wasn't built for it — a city that goes quiet and slightly baffled when it snows, where no one has the right coat, where the cold feels like an intrusion. Not Scandinavian winter, which is competent and expected. Winter as a mild disaster.
Why it works: The contrast between the experience of winter and the "proper" northern European version creates a specific emotional texture — not majestic, not bracing, but awkward and slightly absurd.
Expect: Some British folk songs about weather, certain Beirut tracks, Australian winter indie.
Prompt 42: Music for reading a letter from someone who is no longer alive
Not grief music, not memorial music — the specific texture of the moment of reading words written by someone who didn't know they were writing for the last time. Chamber music, perhaps, or solo piano. Something that holds the past and present simultaneously without resolving toward either.
Why it works: This is a specific, named emotional situation that has very precise tonal requirements — music that holds temporal contradiction rather than resolving it.
Expect: Arvo Pärt, some Satie, Max Richter's more restrained work, certain Keith Jarrett solo recordings.
Prompt 43: The opening scene of a thriller set in a city you've never been to
Music that establishes a city — its rhythm, its social texture, its danger — without showing you around. The kind of score that makes you feel the specific geography of somewhere unfamiliar before a single exterior shot. Probably electronic, probably rhythmic, but oriented around a city's character rather than generic tension.
Why it works: This prompt asks for music that does geographic and cultural characterization work, which is a specific compositional function that both film scores and certain electronic artists specialize in.
Expect: Some Cliff Martinez work, parts of the Tron Legacy score, Nicolas Jaar's more atmospheric tracks.
Section 7: The Personal & Specific Prompts
The most interesting category. These prompts work because they describe experiences that are so specific they feel almost private — and yet when the AI finds the right music, you realize other people have been in exactly that place before.
Prompt 44: Songs my dad would have played if he were cooler
Music my father's generation should have been listening to — the records that were available in the late 1970s and early 1980s that most dads missed because they were watching the game or not paying attention. The hipper version of the standard boomer canon. The music that was playing at the good parties they didn't go to.
Why it works: This is a persona-construction prompt — the AI has to build a character (a cooler, more culturally attuned version of a 50s-born parent) and then find their soundtrack.
Expect: Talking Heads, Television, early Blondie, some Television, Patti Smith's more accessible work.
Prompt 45: Music that sounds different on terrible headphones
Songs that were produced or mixed with an awareness of how they'd sound on cheap speakers or bad earbuds — music where the bass is so embedded in the mid-range that it survives compression and low-end loss. Lo-fi, hip-hop, certain soul and R&B records. The kind of music that sounds better on a phone speaker than on a studio monitor.
Why it works: This is an unusual technical specification that points toward a specific production philosophy — music made for the real conditions in which most people listen.
Expect: Noname, some Soulquarians output, MF DOOM's more compressed mixes, certain Dilla beats.
Prompt 46: The album you bought because of the cover and it turned out to be correct
Music that delivers exactly what its artwork promised — albums where the visual and sonic aesthetics are perfectly aligned, where opening the record is exactly like opening a door into the sound you expected. The ones where you could hear the album from the cover before you played it.
Why it works: This is a meta-prompt about the relationship between visual and sonic aesthetics, which an AI can interpret as music with strong visual identity and a coherent aesthetic world.
Expect: Darkside's "Psychic," Talk Talk's "Spirit of Eden," certain Blue Note covers that delivered.
Prompt 47: What you listen to on the train home after you've been fired
Not angry, not sad — the strange calm of having had something end that you didn't want to end. A slight unreality. The commute that you've done a hundred times looks different today. Music that matches altered perception — familiar things seen from an unfamiliar angle.
Why it works: The emotional specificity (post-shock, altered perception, familiar route looking different) is precise and unusual. It points toward a particular kind of music that operates on slight disorientation.
Expect: The National's slower moments, Low, some early Radiohead album tracks.
Prompt 48: Songs that make you feel like you're watching yourself from outside
Music that creates mild dissociation — a sense of observing your own life from a slight remove. Not alienating, not distressing, but the specific state where you become briefly a character in your own story. Certain electronic music does this, some minimalist composition, late-night pop.
Why it works: Dissociation as an aesthetic experience is well-documented in music criticism, and this prompt names it precisely enough that an AI can identify music that critics have associated with that state.
Expect: Portishead, some Grouper work, certain Beach House tracks, Talk Talk's final albums.
Prompt 49: The last track you want to hear before a difficult conversation
Music for the five minutes before you have to say something hard — the waiting room music for emotional labor. Something that grounds rather than inflates, that brings you back to yourself rather than taking you somewhere else. Not motivational. Not calming in a patronizing way. Just present.
Why it works: The functional requirement here — "grounds rather than inflates" — is specific enough to rule out most of the obvious categories (not meditation music, not pump-up music) and point toward something more grounded.
Expect: Bill Callahan, some Gillian Welch, early Neko Case, certain John Martyn recordings.
Prompt 50: Songs I would have liked before I knew what music was
The music that, if you'd heard it before you developed any taste or context or reference points, would have shaped you differently. Songs so structurally simple or emotionally direct that they bypass everything you've learned about what good music is supposed to sound like. The ones that work on you without permission.
Why it works: This is a deliberately paradoxical prompt — it asks for music that operates below the level of critical listening, which is a real category. Music that bypasses taste and hits something more fundamental.
Expect: Certain Roy Orbison recordings, early Dolly Parton, some Burt Bacharach songs, "The Tracks of My Tears."
How to Use These Prompts
The prompts above are not magic words. They are a method.
The method is: describe the world around the music, not just the music itself. Give the AI a time, a place, a person, a function, a constraint. The more precise the world you describe, the more interesting the playlist that emerges.
Here is how to get the most out of them in practice.
Start with one section and stay there. If you're in a place-and-time mood, use three or four of those prompts back-to-back and compare the results. The differences between "Lisbon kitchen" and "Buenos Aires milonga" will tell you something about how the AI is interpreting your cultural anchors.
Modify the specifics, keep the structure. "A 1997 dial-up modem ritual" works because it has a year, an activity, and an implied emotional register. You can swap the year to 2003, the activity to "waiting for a download to finish," and the register to "nervous energy before results." The structure is what matters.
Refine iteratively. If the first playlist isn't quite right, identify what's wrong and add a constraint. Too upbeat? Add "nothing that would interrupt a quiet room." Too polished? Add "recorded before the studio era could sand off the edges." Too mainstream? Add "specifically not the famous examples of this genre."
Save what works to Spotify. Playgen generates playlists that can be saved directly to your Spotify library. When a prompt produces something excellent, save it — both the playlist and the prompt text. The prompt is as valuable as the playlist it generated.
Use the sections to navigate your own taste. If the cinematic/vibe section resonates and the era/scene section leaves you cold, that's information about how you listen. Some people curate by feeling; others curate by history. The prompts that feel most natural to you are telling you something about your own relationship to music.
The worst outcome is to copy a prompt verbatim, get a mediocre result, and conclude that the tool doesn't work. The tool works. The prompt is the work. Write it like you mean it — like you're describing the playlist to someone who cares enough to get it right.
The best playlists feel like they were made for a specific moment that hasn't happened yet. The ones in your library that you return to years later because they capture something exact and irreproducible. Those playlists always started with a more specific question than "chill vibes."
Go ask a better question. Playgen will answer it.
Generated playlists are saved directly to your Spotify account. Playgen is free to try — no credit card required for your first three playlists.