Aurora, music player Data sheet Rev. V3.2.0, 25 Oct 2026 Apache-2.0

A music player for the files and servers you already own, with a signal path you can inspect.

For Android phones and tablets, Windows 10/11 and Linux beta. Free and open source. Every processed sample runs through a 64-bit engine, and the Signal Path report says whether your samples reached the output untouched, were changed, or can't be verified from where Aurora sits.

Download for Android APK · 132.3 MB All platforms and requirements
Part
Aurora
Revision
V3.2.0, 25 Oct 2026
Runs on
Android 8.0+ (64-bit), Windows 10/11 x64, Linux x86_64 beta
Not made for
macOS, iOS
Licence
Apache-2.0
Checked by
about 1,790 automated tests

Functional block diagramone track, source to DAC

Signal Path verdict Samples preserved

Driver and DAC behavior beyond the exclusive WASAPI stream are not verified

Seven sources feed one merged library and a decoder: files, Navidrome or Subsonic, Jellyfin and Plex blend into it, and YouTube Music, Green Music App and SoundCloud keep their own views, on Android, Windows and Linux (beta). A switch sends the decoded samples either straight to the output (Direct) or through the 64-bit DSP rack (Processed). Outputs are a USB DAC on Android, WASAPI exclusive on Windows and ALSA hw: on Linux (beta); the operating system mixer is not in the path. Files on device Navidrome · Subsonic Jellyfin Plex YouTube Music Green Music App SoundCloud SOURCES own views Direct: decoded samples, untouched to headphones Library merged sources best copy kept Decoder FFmpeg 64-bit float DSP rack binary64 · 16 stages EQ, dynamics, convolution Output USB DAC · Android WASAPI excl. · Windows ALSA hw: · Linux beta DAC OS mixer not in the path FLAC 96 kHz 24-bit P D Seven sources feed one merged library and a decoder: files, Navidrome or Subsonic, Jellyfin and Plex blend into it, and YouTube Music, Green Music App and SoundCloud keep their own views, on Android, Windows and Linux (beta). A switch sends the decoded samples either straight to the output (Direct) or through the 64-bit DSP rack (Processed). The operating system mixer is not in the path. Files on device Navidrome · Subsonic Jellyfin Plex YouTube Music Green Music App SoundCloud own views 7 sources, desktop too Direct: untouched to headphones Library merged sources best copy kept Decoder FFmpeg 64-bit float DSP rack binary64 · 16 stages EQ, dynamics, convolution Output USB DAC · Android WASAPI excl. · Windows ALSA hw: · Linux beta DAC OS mixer not in the path FLAC 96 kHz 24-bit P D
Figure Direct sends decoded samples to the output as they are; Processed runs them through the 64-bit rack first. Verdicts and reasons are worded as the app words them.

Features

  • 7 sources merge: files on the device, Navidrome or Subsonic, Jellyfin and Plex into one library; YouTube Music, Green Music App and SoundCloud join with their own views
  • 64-bit processing on every processed path, edited live without clicks
  • 16 stages in the processing rack, built from 18 stage types; 64 bands per Equalizer stage, 256 per rack
  • 5,057 measured headphones and in-ears, and 1,083 speakers, searchable offline (Android)
  • 384kHz PCM to a USB DAC on Android; WASAPI exclusive on Windows; ALSA hw: on Linux beta
  • DSD1024 native or DoP over USB on Android, from DSD64; DSD64–256 verified on a FiiO KA13‡
  • 20ms analysis frames for Follow vocals, which times lyrics to the singing (Android)
  • 31 visualizer styles, 4 interface styles, 17 languages

Description

Aurora started in June 2026 as an Android player. It now runs on Windows and, in beta, on Linux, sharing its library, server and sound-processing code across all three.

Everyday listening needs none of the audio settings. The advanced parts are there when you want them, and the app says plainly when something is experimental or off by default.

‡Native DSD64–256 and DoP64–128 are verified on a FiiO KA13. Higher rates and other DACs are experimental and off by default.

Files and servers merge into one library, and streaming tracks play your own copy when you have one

Platforms
Android, Windows and Linux beta: every source.
Default
Merge all sources is off until you turn it on.
Since
V1.1.0. YouTube Music merges since V2.5.0, Plex since V2.6.1, Green Music App since V3.0.0, SoundCloud since V3.1.0.

Sign in to as many sources as you like. Files on the device, Navidrome or any Subsonic server, Jellyfin and Plex blend into one library, where copies within ±4 s become one row that plays the best of them. YouTube Music, Green Music App and SoundCloud keep their own Home, Library and Search, and play a copy you own when one matches.

Sources comparedbest audio Aurora gets, by where the music lives

Sources compared, capability by capability
Capability On this device V1.0.0 Android, Windows and Linux beta Navidrome or SubsonicV1.0.0 Android, Windows and Linux beta Jellyfin V1.0.0 Android, Windows and Linux beta Plex V2.6.1 Android, Windows and Linux beta YouTube Music† V2.4.0 Android, Windows and Linux beta Green Music App†‡ V3.0.0 Android, Windows and Linux beta SoundCloud† V3.1.0 Android, Windows and Linux beta
Sign-in NoneUser name, passwordUser name, passwordPlex tokenOptional; Google, in the appIts own page, in the appOptional; its own page
Best audio The file itselfOriginal file§Original file§Original file§Opus or AAC, lossyYouTube match, lossyAAC 160 kbps
Changes you make Likes, playlists, tagsStars, playlistsFavourites, tags, playlistsRatings, playlistsLikes, playlists, followsLikes, playlists, followsLikes, playlists, follows, comments
Lyrics from LRCLIBServer, syncedServer, word-timedServer, LRCLRCLIBIts own, line-syncedLRCLIB
Extras Tag editor, AcoustIDFolders, radioFolders, similar-songs radioShelves, station radioMusic videos, liveCanvas, its own radioTimed comments
Downloads Already thereOriginal or MP3§Original or MP3§Original or MP3§Opus or AAC, as isThe YouTube matchOne AAC file
Plays reported‖ NoneScrobblesPlayback sessionsTimelineYouTube historyNoneSoundCloud history
In the merge Library, always inLibraryLibraryLibraryOwn viewsOwn views, one accountOwn views
Desktop app Yes, music foldersYesYesYesYesYesYes

Ferry Lines on YouTube Music plays your FLAC 96 kHz 24-bit from the phone. The like stays on YouTube Music; a play from your Navidrome copy would scrobble to Navidrome, and one from the phone reports to no service.

9 rows × 7 sources
FigureMarks show the best audio Aurora gets from each source; arrows show a streaming track playing your own copy. Select a row name to mark which sources have it, or a source to read about it. The Quiet Engines are invented for this page.

†Unofficial integrations: a change at the service can break them.

‡We don't print this service's name. Its audio never comes from the service: a copy you own plays, otherwise a matching YouTube recording, which can be another take.

§The original file unless you set a cap: 128–320 kbps on Wi-Fi or mobile data, or about 96 kbps with Data saver on mobile data, and the server sends MP3 instead. Downloads offer the same choice.

‖Last.fm and ListenBrainz scrobble plays from every source except radio and podcasts.

Extension libraries (Android, beta) are read-only and never merge; internet radio and podcasts sit beside the library. How the merge works, step by step, and full data sheets for YouTube Music, Green Music App and SoundCloud: full specification.

How do you listen? Aurora asks this first, in any of its 17 languages.

English. Setup 1 of 4, as Aurora asks it. One of 17 languages, about 2,820 strings each.

  1. Everyday Listening

    Keep the essentials easy to find.

    On this page: Library Lyrics Mix Listening Look

  2. Advanced Audio

    Show detailed output and processing controls.

    On this page: Sound Output the block diagram above

Your answer shapes the setup checklist that follows.

Arabic and Urdu are translated in full; the app's layouts don't mirror right to left yet.

Aurora's first-run setup, step 1 of 4, asking How do you listen?, with the answers Everyday listening and Advanced audio.
Setup asks this first, in the language you picked. Temporary capture: Android emulator, V2.9.0 setup.

Word-by-word lyrics, timed to where the voice actually sings

Word by word: Android, Windows and Linux beta, the default · Follow vocals: Android, off by default · Since V2.9.0

Most synced lyrics only mark where a line starts. Aurora fills each word from the song’s own syllable pace, and Follow vocals goes further: it analyses the track on the phone in 20 ms frames, watches the centre-panned 180–3,800 Hz band where a lead voice sits, and finds where the singing starts, breathes and stops. A dynamic-programming aligner then puts each word break on a pause or a new note, and avoids breaking inside a held one.

Operating conditions
Parameter Value Unit
Analysis frame 20 ms
Vocal band 180–3,800 Hz
Drum rejection 9 frames
Singing threshold Otsu
Break rewarded at ≥ 350 ms

Timing diagramone line of Ferry Lines, 5 s

Ferry lines are folding up the evening air nl · on device‖ De veerlijnen vouwen de avondlucht op

5 s of sound made in your browser: a centred voice over a pad panned wide. This page plays through your browser's mixer, so it can't be bit-perfect itself.
Figure One line of ‘Ferry Lines’, an invented song. Thin spikes are drum hits; a 9-frame median throws them out before the threshold is set. Aurora’s test suite checks the aligner against synthetic singers like the one behind the button.

Otsu separation 6.92; below 1.6 the estimate stays

‖ On-device translation with ML Kit. Since V3.0.0. Off by default, Android only. It translates into the app’s language, skips lines already in it, and downloads each ~30 MB language pack once, over Wi-Fi.

Corrections fitted on the device, from real measurements to a target you pick

Platforms
EQ, rack, squig.link AutoEQ, measurement projects, output binding, preset rules, A/B and ABX: Android, Windows, Linux beta. Measured preset index: Android.
Default
The quick EQ. Rack mode is opt-in.
Since
64-band Equalizer stages in V2.0.0; the Simple and Advanced editor with a draggable graph in V2.9.0 (issue #8).

Start from the bundled offline index of 5,057 measured headphones and in-ears and 1,083 speakers (Android), search squig.link for a live in-ear measurement, or import your own REW, Squig, AutoEq or Wavelet file. Aurora fits peaking filters against Harman, diffuse-field, JM-1 or your own target, sets a negative preamp so boosts can't clip, and can bind the result to the output it was made for, down to a USB DAC's serial number.

Frequency responseinvented IEM, 711 coupler, 20 Hz–20 kHz on a log axis

Made in your browser. This page plays through your browser's mixer, so it can't be bit-perfect itself. Web Audio has 8 filter types to the app's 14; these shelves and peaks match the fit exactly.

Drag a filter, or use the arrow keys: left and right move it by a twelfth of an octave, up and down change its gain by 0.5 dB, and Shift moves four times as far. Page Up and Page Down change a peak's Q. Home and End go to the gain limits.

Error, RMS 20 Hz–10 kHz

Error 3.98 dB before correction, 0.30 dB after 10 filters.

Fitted filters for the Harman IE 2019 target
FilterTypeFcGainQ
1 LSC 105 Hz +6.3 dB 0.71
2 HSC 10 kHz +2.3 dB 0.71
3 PK 2.34 kHz +6.6 dB 1.35
4 PK 6.03 kHz −6.6 dB 4.52
5 PK 8.53 kHz +5.6 dB 5.28
6 PK 182 Hz +1.5 dB 0.79
7 PK 752 Hz −1.8 dB 1.74
8 PK 388 Hz +1.0 dB 2.09
9 PK 15.5 kHz +2.4 dB 3.51
10 PK 9.68 kHz +1.5 dB 6.00
Preamp−6.6 dB

Equalizer APO text

Preamp: -6.6 dB
Filter 1: ON LSC Fc 105 Hz Gain 6.3 dB Q 0.71
Filter 2: ON HSC Fc 10000 Hz Gain 2.3 dB Q 0.71
Filter 3: ON PK Fc 2342 Hz Gain 6.6 dB Q 1.35
Filter 4: ON PK Fc 6032 Hz Gain -6.6 dB Q 4.52
Filter 5: ON PK Fc 8534 Hz Gain 5.6 dB Q 5.28
Filter 6: ON PK Fc 182 Hz Gain 1.5 dB Q 0.79
Filter 7: ON PK Fc 752 Hz Gain -1.8 dB Q 1.74
Filter 8: ON PK Fc 388 Hz Gain 1.0 dB Q 2.09
Filter 9: ON PK Fc 15540 Hz Gain 2.4 dB Q 3.51
Filter 10: ON PK Fc 9682 Hz Gain 1.5 dB Q 6.00
FigureThe measurement and both targets are invented approximations, fitted in your browser with a simplified version of the method: shelves at 105 Hz and 10 kHz first, then peaking filters where the error is largest, trusting the coupler less above 9 kHz. Drag a filter to edit it.

In the app, every processed path runs on 64-bit float samples, and a setting changed mid-song crossfades to the new effect graph at a block boundary, so there's no click.

Absolute maximum ratings
ParameterValueNotes
Processing rack
Stages in the processing rack 16 run top to bottom, reorderable, each with wet/dry and bypass
Stage types 18
Bands per Equalizer stage 64
Bands across the rack 256 512 biquad sections in total
Inputs per stage 4 stereo, left, right, mid, side, M/S encode and decode; branches delay-compensated up to 2 s
Filter types 14
Equalizer and corrections
Bands in the quick EQ 12
Bands in an AutoEQ or squig.link correction 12 the squig.link generator makes up to 10

Filter typesmagnitude at 48 kHz, +12 to −24 dB, 20 Hz–20 kHz

  1. Peak bell, gain and Q
  2. Low shelf gain below Fc
  3. High shelf gain above Fc
  4. Low pass 12 dB/oct
  5. High pass 12 dB/oct
  6. Band pass Q sets the width
  7. Notch Q sets the width
  8. All pass phase shown
  9. Tilt two shelves around Fc
  10. Butterworth LP 12–48 dB/oct
  11. Butterworth HP 12–48 dB/oct
  12. Linkwitz–Riley LP 24 or 48 dB/oct
  13. Linkwitz–Riley HP 24 or 48 dB/oct
  14. Custom biquad 5 coefficients
FigureComputed in your browser from the same biquad formulas as the app's engine. Each band of a rack Equalizer stage can use any of the 14. The plot shows the Peak filter.

Bit-perfect paths on the phone, on Windows and on Linux beta

USB Direct on Android · WASAPI exclusive on Windows · ALSA hw: on Linux beta · Off by default; switched on per device · Since V2.1.0 (USB), V2.7.0 (Windows), V2.8.0 (Linux beta)

On Android, Aurora brings its own user-space USB Audio Class driver, so samples reach a DAC without passing Android's mixer. In Direct mode they are left alone: the volume keys move the DAC's own volume, and its clock is read back to check it runs at the rate asked for. On Windows the same job is done by WASAPI exclusive mode, on Linux beta by a direct ALSA hw: device; on both, Aurora's volume sets the DAC's own volume when it has one. When a device refuses a format, Aurora falls back and says why.

Output characteristics Rev. V3.2.0
Parameter Conditions Min Typ Max Unit
Engine, all platforms
Internal sample format Every processed path not specified binary64 not specified 53-bit significand
Output rate policy Follow source, fixed, or highest compatible 44.1 not specified 384 kHz
Android, USB Audio Class 1 and 2
PCM rate Isochronous, async feedback; DAC clock read back and checked not specified not specified 384 kHz
Word length Mono or stereo 16 not specified 32 bit
DSD, native or DoP Direct mode only* DSD64 not specified DSD1024 2.82–45.16 MHz
Desktop, exclusive mode
WASAPI exclusive, Windows 16, 24, 24-in-32 or 32-bit integer§ 44.1 not specified 384 kHz
ALSA hw:, Linux beta Pick the card itself, not the PipeWire entry 44.1 not specified 384 kHz
Where the volume is applied
Path Volume control Samples
Android, Direct USB The DAC's hardware volume, following the phone's media volume through its USB feature unit untouched
Android, Processed USB Software volume in the 64-bit engine modified
Windows or Linux beta, exclusive The DAC's own volume when it has one; otherwise software volume, which does nothing at 100% untouched with DAC volume, or at 100%
Android, system output speaker, Bluetooth, USB through Android Android's mixer unknown; on Android 14+ Aurora can ask for the system's bit-perfect USB mixer

*A DAC's advertised DSD rate alone doesn't guarantee it works. For experimental rates, disconnect your headphones until the DAC confirms DSD mode and start at low volume.

§Driver and DAC behaviour after the WASAPI or ALSA stream is reported as unknown, not as verified. Software volume below 100% (when the DAC has no volume control of its own), ReplayGain, crossfade and DSP all change samples, and Signal Path says so.

All 16 characteristics, with conditions: full specification.

Signal Paththe app's report format

Signal Path verdict Samples preserved

  • Driver and DAC behavior beyond the exclusive WASAPI stream are not verified
Aurora Signal Path · diagnostic v1
State: Active
Sample preservation: Preserved
- Driver and DAC behavior beyond the exclusive WASAPI stream are not verified

Source: flac · FLAC · 2764800 bit/s
Format: 96000 Hz · 24-bit · 2 channels · integer PCM
Evidence: FFmpeg stream probe; upstream transcoding history may be unknown

Decoder: FFmpeg flac
Format: 96000 Hz · 64-bit · 2 channels · float PCM
Evidence: FFmpeg decoder output converted to binary64 stereo

Processing: No active Aurora sample-processing node
Evidence: Current chain configuration and runtime processing state

Resampling: Following source rate
Evidence: Configured chain rates

Output: Exclusive WASAPI stream · 24-bit integer in 32-bit container
Format: 96000 Hz · 24-bit · 2 channels · integer PCM
Evidence: Open WASAPI stream format

Device: Selected output · DAC hardware volume −12.0 dB
Evidence: Active WASAPI endpoint; hardware format and downstream processing unknown

Latency: Rack: 0 frames · SRC: 0 frames · Tail: 0 frames
Evidence: Compiled processor delay; output buffer and hardware latency are not included

Hardware behavior beyond the reported output boundary is not verified.

Figure An abridged example in the app's report format; the values are illustrative, the wording is the app's. A copied report never includes titles, URLs, account or device IDs.

Sample tapewords compared one by one

Windows · WASAPI exclusiveDirect

compared 12 words differ 0

Figure 24-bit words as decoded and as handed to the output. With processing on, almost every word changes, which is what ‘Samples modified’ means. Generated in your browser.

DSDa 1-bit stream, and what a low-pass filter recovers from it

Figure DSD is a stream of single bits whose density follows the waveform. By default Aurora converts it to 176.4 kHz PCM so your EQ still applies. In Direct USB mode it can send it untouched, as DoP or native DSD.
DSD support by rate
Rate Carrier, MHz To PCM DoP Native
DSD64 2.8224 yes tested tested
DSD128 5.6448 yes tested tested
DSD256 11.2896 yes experimental tested
DSD512 22.5792 yes experimental experimental
DSD1024 45.1584 yes experimental experimental

Tested means verified on the FiiO KA13, the reference DAC. Experimental rates sit behind an off-by-default switch. DSF and DFF, mono or stereo; DST-compressed DFF at DSD64.

Mixthe join from Ferry Lines into Keepsake, 32 beats

Figure Each lane shows the raw waveform faintly and the gain-applied one solid. The three fade shapes are the app's own curves: smoothstep, linear and equal-power.

Ferry Lines and Keepsake are invented for this page.

Mix plays an album or playlist as one continuous set, and works out every join on the phone

Tap Mix and Aurora keeps your track order but treats each change like a DJ would. Tempos within 6% are matched and the incoming cue snaps to the beat; on those beat-matched joins the bass of one track hands over to the next instead of doubling up. Similar keys get 16 beats, others 8. Loudness is levelled toward the set's median by cutting, never by boosting.

Platforms
Mix, Mix Studio, stems and crossfade: Android, Windows and Linux beta.
Default
Mix is a button on every album and playlist. Crossfade is off.
Not with
Exclusive USB output, or while casting.

Mix Studio opens any mix as a multitrack arrangement of up to 8 layers, each clip with its own cues, fade, gain, pan, three-band EQ, tempo (0.5–2×) and pitch (±12 semitones). Stems play the vocals or the backing alone: the UVR MDX-Net Voc_FT model separates them on the phone after a one-time 64 MB download, and the audio never leaves the device.

Your listening, counted on your device and kept there

Statistics and recaps for every day, week, month, year and all time are built in. A play counts after 30 seconds, or half of a shorter song. Once you set up detailed statistics, Aurora also records when and how you listen, and what share went out lossless, hi-res or bit-perfect. That data sits in storage left out of backups and is never uploaded.

Platforms
Statistics and recaps everywhere. Detailed statistics, places and the map on Android, Windows and Linux beta.
Default
Detailed statistics are off until you set them up.
Since
Recaps in V2.5.0; detailed statistics and the listening map in V2.9.0.

When you listen7 days × 24 hours, average week

Monday
0 min
0 min
0 min
0 min
0 min
0 min
0 min
23 min
21 min
0 min
0 min
0 min
12 min
0 min
0 min
0 min
0 min
10 min
22 min
0 min
11 min
22 min
11 min
0 min
Tuesday
0 min
0 min
0 min
0 min
0 min
0 min
0 min
16 min
32 min
0 min
0 min
0 min
0 min
0 min
0 min
0 min
0 min
7 min
30 min
0 min
18 min
40 min
11 min
0 min
Wednesday
0 min
0 min
0 min
0 min
0 min
0 min
0 min
5 min
31 min
0 min
0 min
0 min
6 min
0 min
0 min
0 min
0 min
10 min
27 min
1 min
12 min
34 min
11 min
0 min
Thursday
0 min
0 min
0 min
0 min
0 min
0 min
0 min
9 min
30 min
0 min
0 min
0 min
7 min
0 min
0 min
0 min
0 min
8 min
32 min
0 min
19 min
43 min
10 min
0 min
Friday
0 min
0 min
0 min
0 min
0 min
0 min
0 min
25 min
24 min
0 min
0 min
0 min
7 min
0 min
0 min
0 min
0 min
9 min
19 min
0 min
17 min
35 min
25 min
6 min
Saturday
0 min
0 min
0 min
0 min
0 min
0 min
0 min
0 min
0 min
0 min
13 min
19 min
5 min
0 min
0 min
2 min
6 min
0 min
0 min
0 min
28 min
10 min
29 min
30 min
Sunday
0 min
0 min
0 min
0 min
0 min
0 min
0 min
0 min
0 min
0 min
6 min
36 min
22 min
0 min
0 min
1 min
17 min
0 min
0 min
0 min
33 min
5 min
22 min
31 min
Day 00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00
Slice
Sat–Sun 20:00–23:59
Listening
3h 8min
Plays
41
Average session
57min
Top artist
The Quiet Engines
Top genres
ambient, folk, electronic
Outputs
USB 70% · wired 30%

Audio path · whole week, share of listening time

  • Lossless 86%
  • Hi-res 39%
  • AutoEQ 64%
  • Bit-perfect 19%

Outputs, whole week: wired 41% · USB 30% · Bluetooth 25% · phone speaker 4%

Figure An average week of listening, hour by hour, filled on the app's square-root scale. The margins total each hour and each day, with the selected slice's part in rose. Selecting a block of days and hours reads that slice, as the app's detailed statistics do.

Example data for a fictional listener. Shares are weighted by listening time.

A September recap picture in the Midnight style: minutes listened, top artists and top songs, with placeholder names.
A September recap picture in the Paper style: minutes listened, top artists and top songs, with placeholder names.
A September recap picture in the Aurora style: minutes listened, top artists and top songs, with placeholder names.

Recap picture1080 × 1600

Midnight.Paper.Aurora. Any recap saves as a 1080 × 1600 picture. Temporary capture: Aurora's recap renderer.

Places and the map, Private Session, scrobbling and the recap inbox: full specification.

Four interface styles, and album pages that take the cover's colour

Aurora, Retro hi-fi, Aero and Liquid Glass each have their own controls, surfaces, type, corners and colours. In the Aurora style, album, artist and playlist pages take their colour from the artwork, picked in the order shown here: vibrant first, then light vibrant, dominant and muted, lifted or darkened if it would vanish against the page.

Platforms
Styles, modes, accents and the visualizer everywhere. Launcher icons on Android; live glass refraction on Android 12+.
Default
Aurora style, dark mode, Rose accent.
Since
Album tint in V2.7.0; live Liquid Glass in V3.0.0.

Specimen · the Now Playing screen, redrawn for each style Liquid Glass · drag the lens along the tab bar, or focus it and use the arrow keys

Sampled from the cover · pick order

  1. Vibrant none
  2. Light vibrant #F87858 picked
  3. Dominant #102028
  4. Muted #907868

Plate #F2DCD0 · text #AD3311 · 4.89:1 Plate #382020 · text #F87858 · 5.60:1

  1. Vibrant #30A878 picked
  2. Light vibrant none
  3. Dominant #082820
  4. Muted none

Plate #D6E2D4 · text #006C48 · 4.84:1 Plate #1C2725 · text #30A878 · 5.12:1

  1. Vibrant #6888B8 picked
  2. Light vibrant #90C0F8
  3. Dominant #081030
  4. Muted #5878A0

Plate #DEDEDD · text #415F8C · 4.81:1 Plate #24232E · text #7191C2 · 4.82:1

Interface style · with its launcher icon on Android
Accent · Aurora style

Rose #FF2E7E · plus a custom colour or Material You

Album tint

Sampled at build time from these covers with a port of the app's rule. The covers are drawn from each album's own tracklist: one light per track on Harbour Lights, one pane per track on Glasshouse Weather.

Real-time analyser5 of the app's 31 styles

Silent preview, drawn from the test signal's score

Made in your browser. This page plays through your browser's mixer, so it can't be bit-perfect itself.

Figure In the app the visualizer reads Aurora's own pipeline, so it works in bit-perfect mode too, and it never asks for the microphone. Music Haptics (Android) reads the same pipeline to tap the phone along with the drums and bass. All 31 styles: full specification.

* Live refraction in Liquid Glass, where the tab bar and mini player bend and blur what scrolls beneath and a lens follows your finger, needs Android 12 or newer. On Windows and Linux (beta), Glass is a static glass-style palette.

System, Light, Dark or AMOLED modes, typefaces, text size, corners and launcher icons: full specification.

Eighteen releases in four months, from one developer and one regular contributor

nessli420, creator · shuja, Linux beta port and packaging, ALSA output, 14 translations, YouTube downloads, USB Audio Class 1 · Apache-2.0

Commits per day13 Jun to 25 Oct 2026 · releases marked

commits release

Arrow keys move one day, Page Up and Page Down one week, Shift with an arrow to the next day with commits or a release.

Figure Commits per day, 13 June to 25 October 2026, with releases marked. Fifteen releases, V2.0.0 to V3.2.0, landed in 38 days.

308 commits by two people · git 220c89b

Report to fix

Hours from an outside report on GitHub to a fix on main.

  1. #12 Last.fm fix 2 hours

  2. #10 DAC hardware volume 4 hours

  3. #14 Jellyfin 12.2 sign-in 8 hours

  4. #5 Pull-to-refresh 12 hours

  5. #2 Profiles, artist separators 13 hours

  6. #9 Animated covers 16 hours

  7. #7 USB regression 17 hours

  8. #6 YouTube downloads 19 hours

  9. #8 Equalizer redesign 2.4 days

Median 13 h across all nine.

Errata

  • Windows installers aren't code-signed, so SmartScreen may ask before the first launch.
  • The Gapless switch is saved but not read. Playback is gapless either way.
  • Scrobbles that fail offline aren't retried later, for Last.fm or ListenBrainz.
  • Arabic and Urdu are translated, but layouts don't mirror right to left yet.
  • On GNOME the tray needs AppIndicator; without a keyring, passwords last until Aurora closes.
  • Lyrics embedded in local files and .lrc sidecars aren't read yet; local tracks use LRCLIB.
  • A backup made with Include sign-ins holds login tokens and API keys in plain text. Keep it private.

Found another? Report it

On GitHub now

Connecting to GitHub

Downloads, all platforms
—
  • Android —
  • Windows —
  • Linux beta —
Stars
—
Commits
—
Releases
—

Latest release

All releases

Recent commits

All commits

Live numbers from GitHub load with JavaScript. The repository, its releases and its commits are public:

Download V3.2.0

Released
25 October 2026, six packages. Each link downloads that file from the V3.2.0 release on GitHub.
Updates
Android checks GitHub, verifies the download and updates itself; Windows downloads and runs the new MSI from Settings → About. On Linux beta, install the new package.

Packagesa hint from your browser Phones and tablets share one APK. Windows has two installers, Linux beta three packages.

Packages for V3.2.0. Each file link downloads it from that release on GitHub.
Platform File Size Requires Status
Android your system phones and tablets Aurora.apk 132.3 MB Android 8.0 or newer, 64-bit Stable
Windows your system installer Aurora-3.2.0.exe 204.5 MB Windows 10 or 11, x64. Java is bundled. Stable, not code-signed
Windows your system MSI, replaces older installs Aurora-3.2.0.msi 203.6 MB Windows 10 or 11, x64 Stable, not code-signed
Linux beta your system Debian, Ubuntu and derivatives aurora-deb.zip 187.0 MB x86_64, glibc 2.35 or newer beta
Linux beta your system Fedora, openSUSE aurora-rpm.zip 189.8 MB x86_64, glibc 2.35 or newer beta
Linux beta your system any distribution aurora-appimage.zip 199.8 MB x86_64, FUSE beta

Linux install commands, requirements and building from source: Download page. There is no macOS or iOS version.