Preamp

2026

Reimagining Winamp for streaming

Year

2026

Role

Self-initiated concept

Timeframe

Ongoing

Team

1 DES

Tools

Figma, Swift, SwiftUI

A wireframe of Preamp with every module open, docked around the main window. A visualizer, an equalizer with a preamp slider and ten bands, sources shown as coloured shapes, output devices, a playlist of real songs with covers and source badges, Ask with a request and a reason under each pick, a live radio station and a list of premium connectors. — How might one small player hold every service you pay for, every station and every file you own? A first look, every module open, still in wireframe.

TL;DR

Preamp is Winamp reimagined for streaming. An unsolicited concept for a player that belongs to the listener again, built on the ideas that made Winamp matter in the first place.

The player became a storefront

Two players defined desktop listening for a generation. foobar2000 was the purist’s tool, fast, modular and endlessly configurable. Winamp was the one people loved. A 275 by 116 pixel window that people skinned, extended and kept open all day. Both did one thing well. You dropped music in, it played, and you made it your own.

foobar2000 still has a devoted following, and its component model shaped this concept too. The foundation is Winamp, because its ideas reached everyone, not only power users. The concept gets a name of its own, Preamp. A preamp sits in front of every source and shapes the signal before it reaches the speakers. That is the job here, for services, radio and files alike.

Desktop streaming apps carry far more music and far more ambition. They also carry a home screen full of promotions, a library that strains past a few thousand songs, and an interface ported from a phone. Plenty of listeners pay for more than one service, keep a folder of files, and still tune into the radio. No single app holds all of it.

This concept asks one question. What if Winamp came back as Preamp, an open platform where every subscription, every radio station and an AI that works for the listener plug into one small, fast, personal player?

The brief I set myself

Keep the classic window. New capabilities sit around it, never inside it.
Treat streaming services as sources, not destinations. The listener connects the subscriptions they already pay for.
Make AI a native capability that answers requests, never a feed that pushes.
Find a business that works without licensing a catalogue.

Assumptions

This is set a few years out. It assumes services let authorized third-party players stream to their subscribers, the way they already allow casting and car integrations. Platforms appear only as colour and shape, wireframed. This is a personal exploration with no affiliation to Winamp’s owners or to any service shown.

Ten patterns listeners push back on

The same frustrations come up again and again in reviews, forums and conversations about desktop music apps.

Two wireframes side by side. On the left, a generic streaming home screen marked with ten numbered pain points. A promoted banner, a Made for you row, a cramped library sidebar with locked playlists, a horizontal carousel arrow, a greyed-out unavailable track, a thin now-playing bar with autoplay, a ghost of another app with the same layout behind it, credits hidden behind an overflow button, and an AI DJ card. On the right, the Winamp concept opens to the track that was playing and the queue it came from. — Left, the common desktop pattern, numbered to match the list below. Right, the concept opens where you left off.

The home screen is a billboard

Promoted releases, podcasts and audiobooks share the first screen with music. It reads as the app working for its business, not for the person who opened it.
The algorithm folds in on itself

Recommendations converge on the same few dozen songs. Smart shuffle feels rigged, because it is.
The library is an afterthought

Saved songs pile into one giant list. Sorting, folders and bulk edits are thin. Collectors suffer most.
Desktop is a ported phone

Horizontal carousels, one window, weak keyboard support and heavy memory use.
Ownership feels fragile

Songs grey out or vanish from playlists without warning.
Playlists are locked in

Years of curation stay behind the service that holds them.
Autoplay takes over

An album ends and the app keeps going with something nobody chose.
Every app looks the same

Dark canvas, album grid, one accent colour. No personality and no control.
Credits are buried

Writers, producers, labels and liner notes sit three taps deep or nowhere.
AI arrived uninvited

Voices talk over the music, picks come with no explanation, and generated filler tracks appear unlabelled.

Every pattern moves control from the listener to the platform. Winamp’s original deal ran the other way. The listener chose the music, the look and the extensions. The player stayed out of the way.

Back to first principles

Winamp did not win on features. It won on a contract with the listener.

Small by default

The player sits beside your work, not on top of it.
Opens where you left off

No home screen. The last track, position and queue come back.
Your library, your rules

Real sorting, smart playlists and bulk tagging, across every source.
Open to extension

Plugins and skins are first-class. The community shapes the product.
Fast and quiet

Native, light on memory, no nags, no promotions.

Every service becomes a plugin

Classic Winamp was a shell with slots. Input plugins decoded formats, DSP plugins shaped sound, visualization plugins drew to the music, and general plugins added anything else. Skins redrew the whole thing. That model maps onto streaming almost one to one.

Classic Winamp's plugin slots, mapped onto streaming. The listener owns the core and the shell.

Input slot

Each subscription, radio directory and local folder is a source. Every source plugs in through one open connector protocol, and the listener connects them once.
Core

One library, one queue and one taste profile. They belong to the listener, not to any source.
DSP slot

EQ, headphone correction, crossfade and loudness, applied the same way to everything.
Intelligence slot

The AI reads the taste profile and writes to the queue. It can be swapped or switched off.
Shell

The skin. The listener decides what the product looks like.

Connectors, in the spirit of MCP

The input slot works the way the Model Context Protocol does for AI tools. One open standard describes what a source can do. Search, stream, sync a library, return credits and artwork. Any service, label, archive or independent developer can ship a connector without asking the player for permission.

The same protocol feeds the intelligence slot. The AI sees every source through it, so a new connector makes recommendations better on the day it ships.

Core connectors stay free. Premium ones cost extra. A hi-res download store, a DJ record pool, a live concert archive or a label’s back catalogue can each charge for their connector, and the player takes a share. The old plugin directory grows into a marketplace.

One library across every source

The same song often lives on several services. The library stores it once, as a canonical track, with every available source attached. The listener ranks sources once, by quality or by service, and the badge on each row shows which one will play.

A playlist wireframe where every row carries a coloured source shape. The playing track opens a card listing four sources for the same song, with the lossless one marked as preferred. A radio row opens a second card showing a live state with no seeking and a Save song button. — One canonical track, many sources. The listener sets the order once, and the badge shows which one plays.

When a service drops a song, the library keeps the entry and falls through to the next source. A track only greys out when nothing the listener has can play it. Playlists stop belonging to a service and start belonging to the person who made them.

Radio stays live

Radio is a different kind of listening and gets its own state. An on-air indicator, no seeking, and the current song shown as it airs. Save song resolves the track against the listener’s sources and drops it into the library.

The tiny window, upgraded

The main window at three times size with seven numbered callouts. One, the context line. Two, the source badge. Three, the quality readout. Four, the time and visualizer. Five, the panel toggles. Six, the command bar hint. Seven, quick panels for radio and files. — Same 275 by 116 footprint. Seven changes, none of which grow the window.

Context line

The scrolling title now says why a track is playing. “New from an artist you play most” or “From your files, unplayed in two years.”
Source badge

A small coloured shape shows where the audio is coming from.
Quality readout

The old kbps and kHz boxes become a plain quality label, so lossless is visible, not assumed.
Time and visualizer

Unchanged. Some things were right the first time.
Panel toggles

EQ and playlist toggles stay. Sources, Ask and Credits join them as panels.
Command bar

One keystroke opens search, requests and actions, so the window never needs a search field.
Quick panels

Radio and files are one click from the main window. Ask lives in the command bar.

Panels snap, the main window stays

New capability arrives as panels on the classic 275 pixel grid. They snap to the main window, to each other and to screen edges, the way the equalizer and playlist always did. The listener builds the player they want, and the main window never grows.

Three modes, one player

Three modes. Mini is a single-line strip under a menu bar icon, playing Kukushka by Kino. Docked is the classic main window with a playlist of real songs snapped below it. Full is a large library window themed by the active skin, with an album grid of real covers, a playlist built by Ask, sidebars for library, sources and playlists, and the transport docked along the bottom. — Mini, Docked and Full. One player, three sizes, the same controls in the same places.

Mini

A menu bar icon and a single strip, in the spirit of the old windowshade mode. For people who want music and nothing else.
Docked

The classic stack with panels snapped as needed. The default.
Full

A large browse window for library work and long sessions, themed by the active skin. The main window docks at its base as the transport, so controls never move.

Keyboard first

A macOS desktop with the Preamp menu bar and the player docked in the lower left. A command palette shows the query rainy sun, with an Ask suggestion to build a queue, the Rainy Sunday playlist, and actions with their keyboard shortcuts. — One keystroke for search, requests and actions.

Every action has a shortcut. The command bar handles search, requests and commands in one place. Media keys, Control Centre and Shortcuts work as expected on macOS.

Skins, with a floor

The same Preamp player in four skins, Graphite, Paper, Classic 2.x and Night Drive. Each carries a badge with its weakest text contrast ratio, from 6.2 to 1 up to 9.1 to 1, all passing WCAG AA. — Four skins, one layout. Every skin shows its weakest contrast ratio, and all four pass AA.

Skins return as the product’s identity and its community. A new skin format adds vector scaling, light and dark variants, and a contrast check that keeps text at WCAG AA. Hit targets keep a minimum size at every scale. Classic skins still load in a compatibility mode. The old double-size option becomes proper scaling for high-density displays.

AI that waits to be asked

The AI is a capability, not a landlord. It answers requests and stops.

The AI answers a request and stops. It never fills a feed autonomously, although it can.

On request

“40 minutes, rainy Sunday, slow and warm” builds a queue. It never fills a feed on its own, although a listener can ask it to.
Only what you can play

Picks come from connected services and local files. It never suggests something the listener cannot hear. Radio stays out of AI queues, since nobody can know what a live station will play next.
Reasons on every pick

One line explains each choice, so the listener can see the logic.
Steer, don’t scroll

“Less like this,” swap a source or regenerate. Each nudge updates the taste profile.
Silent by default

No voice talking over the music.
AI-generated music labelled

Generated tracks carry a tag, and one toggle hides them everywhere.
Private taste

The taste profile lives on the Mac, can be exported, and can be reset.

What success looks like

A concept without targets is a mood board. These are the measures I would hold it to.

A metrics tree. The north star at the top is weekly hours of chosen listening, meaning plays the listener started or steered, with autoplay excluded. Four drivers sit below it, activation, habit, discovery and expression. A row of guardrails sits at the bottom, memory footprint, autoplay share, AI-track complaints and free-to-Pro conversion with churn. — A north star that rewards intent, not time on screen.

North star

Weekly hours of chosen listening. It counts plays the listener started or steered and leaves out autoplay. Time in an app is easy to inflate with a feed. Chosen listening only grows when the player does its job.
Drivers

Activation is a source connected and a first play in the first session. Habit is D30 retention and days listened per week. Discovery counts new-to-you artists that get saved, not merely played. Expression tracks skins installed, made and shared, the loop that turned Winamp into a community.
Guardrails

Memory footprint stays low. Autoplay share stays small. Complaints about AI-generated tracks trend down. Conversion and churn keep the business accountable for value.

The business

The player is free with every source, the library, radio and skins. A Pro tier adds the advanced DSP chain, sync across Macs and higher AI limits. Premium connectors are sold individually or bundled into Pro. A skin, plugin and connector marketplace shares revenue with creators. Services could pay for subscriptions that start inside the app. No catalogue licensing, no ads.

Where this could fail

Nostalgia fades

Novelty carries the first week. Retention depends on the library and the AI beating what people already use. Test D30 on listeners too young to remember the original.
Services restrict access

Files, radio and purchases still work without them. The product’s value cannot depend on a single partner.
Seams between sources

Gapless playback and crossfade across different services may not be possible. When a seam exists, the queue marks it instead of hiding it.
Skins break usability

The contrast and target-size floor is part of the format, not a guideline.
Plugins become a threat

Signed plugins and connectors, scoped permissions and a reviewed directory.
Scope creep

One rule. The main window never grows.

What I would test first

A five-second test on the main window. Can people read the context line and the source badge at a glance?
A diary study with listeners who use more than one service, run against a merged library built from their real collections.
A concept test comparing a request-first AI with a feed. Measure trust, steering and saves, not time spent.

Everything open

The main window with every module docked around it, the way Winamp’s windows used to snap together.

Preamp on a macOS desktop with every module open in the Graphite skin. The main window plays Soleil Bleu. Docked around it are a visualizer, an equalizer with a preamp slider and ten bands, sources, output devices, a playlist of real songs with covers and source badges, the Ask module with a reason under each pick, a live radio station and a list of premium connectors. — Fully loaded. The main window and eight modules, one skin, one grid.

Nobody listens like this every day, and nobody has to. Each module is optional and snaps to the same grid, so the player can stay a single strip or grow into a full control room. The skin carries across all of them, and the main window keeps its place at the top left.

A name and a mark

A name that explains the job, and a mark drawn from the hardware it replaces.

A p drawn from the ticks of a gain knob, cut out of the knob itself. Its stem is a volume bar.

Preamp keeps Winamp’s spirit without borrowing its identity. The mark is a lowercase p drawn entirely from the ticks around a gain knob. The bowl is the dial, and the stem is a volume bar that grows heavier as it drops. The whole p is cut out of a disc, the face of the knob, so it reads as one object. Inside the app, the p drops its disc and works on its own, white in the menu bar and in the skin’s accent on the player.

Construction. Every tick sits on a 30° ray between radius 32 and 68. The stem bars hang from the nine o'clock tick and thicken by three units each. The disc centres on the whole p at its optical centre.

Turning it up. The icon lights from the bottom of the stem, through the dial, then settles back.

Amber comes from an LCD glowing in a dark room. Graphite and Paper anchor the two default skins, and Signal is the red that Paper uses for its accent.

Interface text is set in SF Pro, so the app feels at home on macOS. Monospace appears only on the LCD, where it belongs.

The voice is quiet, exact and a little warm. It tells you why a song is playing and then gets out of the way.

In motion

PREAMP

02:41

Soleil Bleu

Bleu Soleil & Luiza

New from an artist you play most

LOSSLESS 24/96EQPL

2:414:06

Ask

PREAMP

00:48

Harlem River

Kevin Morby

From your files. Unplayed in 2 years

LOSSLESS 24/96EQPL

0:489:15

Ask

PREAMP

LIVE

Late Hours Radio

Now playing Wildfires by Sault

Live. Save a song and it lands in your library

320 KBPSEQPL

ON AIRLIVE

Ask

PREAMP

01:12

Control

Warhaus

Picked by Ask for a rainy Sunday

LOSSLESS 24/96EQPL

1:123:55

Ask

Every few seconds, the next track from a different source in a different skin. The context line says why each one is playing.

At home on the Mac

Small by default. Preamp sits beside the work, in the menu bar and in the Dock, and stays out of the way.

FAQ

Why Winamp, and why a new name?

Winamp solved the problems today's apps created. A small window, a library that belonged to the listener, plugins and skins that let a community shape the product. Those ideas fit a world of many services better than a world of one. The concept carries a name of its own, Preamp, so it can stand on its ideas rather than on nostalgia.

Has this ever been done?

Parts of it, never all of it. Roon merges local files with Tidal and Qobuz in one library, for a paid audiophile setup. The Sonos app searches every connected service from one home screen, but it exists to control Sonos speakers. Music Assistant, a self-hosted Home Assistant project, links the same track across Spotify, Tidal and local files and plays the best copy. The closest is Winamp itself. In July 2026 it announced a player for the first half of 2027 that brings streaming, local files, radio and podcasts together. Its subscription is Winamp's own, white-labelled from Deezer, so it adds one more catalogue instead of connecting the ones listeners already pay for. Preamp bets on the opposite.

Why not license a catalogue?

Licensing is the cost that sinks independent music services. Listeners already pay for catalogues. The open space is the player that brings them together, so the business sells the experience, not the music.

Why would a streaming service let a third-party player sit between it and its subscribers?

Services already do this for cars, speakers and game consoles, because reach beats control. Preamp gives them heavy listeners they might otherwise lose, subscriptions started inside the app with a referral share, and listening data on their own catalogue. A service that refuses still loses nothing it had. Files, radio and purchases keep the product useful without it.

Where does the money come from if you never own the music?

Four lines, all of them without licensing costs. A Pro tier for the advanced sound chain, sync and higher AI limits. A revenue share on premium connectors. A cut of the skin and plugin marketplace. Referral fees from services when a subscription starts in the app.

Isn't Mac-only too small a market to bother with?

It is a wedge, not the destination. Mac listeners skew toward paying for sound quality and good software, which is the audience a premium player needs first. Native code is also what makes the light and fast promise real. Once retention holds there, Windows is the obvious second platform, since it is where Winamp grew up.

How do you know people want this beyond nostalgia?

Right now, nobody does. That is why the first three tests sit in the case study. The one that matters most is retention among listeners too young to have used Winamp. If they stay past the first month, the idea stands on its own.

What would you cut from the first release?

Everything that is not the core loop. The main window, the playlist, a merged library across two services and local files, and the equalizer. Ask, the skin marketplace, premium connectors and Full mode wait until people prove they come back daily. The context line stays, because it is the clearest difference from every other player.

If the AI never fills a feed, how does discovery compete?

It competes on quality, not volume. Discovery is measured by new artists that get saved, not by how much gets played. The context line explains every pick, Ask answers specific moods, and radio brings in human curation. Passive listening will be lower than in a feed. That trade is the point of the product.

Can one queue mix protected streams from different services, with one equalizer on top?

Not everywhere, and the design does not pretend otherwise. Each service will likely require its own playback SDK, and some may block processing of their audio. Where crossfade or the equalizer cannot reach a source, the queue marks the seam and the equalizer shows that source as bypassed. Partnerships decide how much of that disappears.

How do you match the same song across services?

Through ISRC, the standard code that identifies a recording, which most catalogue APIs already expose. Audio fingerprinting covers releases without one. When a match is wrong, Swap source lets the listener correct it, and that correction improves matching for everyone.

Why native instead of a cross-platform web shell?

Because the promise is light, fast and at home on the Mac. Native code keeps memory low, respects battery life and hooks into media keys, Now Playing, AirPlay and Shortcuts without workarounds. The cost is a separate Windows build later, which is a fair trade for a player that is open all day.

Is this affiliated with Winamp, foobar2000 or any streaming service?

No. Preamp is an unsolicited concept made to explore a design question. Services appear only as colour and shape.

The playlist

Every song in this study, in order of appearance, plus one bonus track. All of them are real monthly picks.

01Soleil BleuBleu Soleil & Luiza
02NasnusaSababa 5 & Yurika Hanashima
03Harlem RiverKevin Morby
04Mum Does the WashingJoshua Idehen
05ControlWarhaus
06Lips Like SugarEcho & The Bunnymen
07IkigaiRobohands
08The Astounding Eyes of RitaAnouar Brahem
09AtomizedAndrew Bird
10Innundir SkinniÓlöf Arnalds
11Space Is Only Noise If You Can SeeNicolas Jaar
12WildfiresSault
13KukushkaKino
14Bésame Mucho (Live)Zoé
15Mumbo SugarArc De Soleil
16It's a Good Day (To Fight the System)BonusShungudzo

The player became a storefront

The brief I set myself

Assumptions

Ten patterns listeners push back on

What they share

Back to first principles

Every service becomes a plugin

Connectors, in the spirit of MCP

One library across every source

Radio stays live

The tiny window, upgraded

Panels snap, the main window stays

Three modes, one player

Keyboard first

Skins, with a floor

AI that waits to be asked

What success looks like

The business

Where this could fail

What I would test first

Everything open

A name and a mark

In motion

At home on the Mac

FAQ

The playlist