diff --git a/.gitignore b/.gitignore
index 8f9bbdae..27c02da3 100644
--- a/.gitignore
+++ b/.gitignore
@@ -35,4 +35,5 @@ AGENTS.md
 *.test
 *.wasm
 *.ndp
-openspec/
\ No newline at end of file
+openspec/
+go.work*
\ No newline at end of file
diff --git a/adapters/gotaglib/end_to_end_test.go b/adapters/gotaglib/end_to_end_test.go
new file mode 100644
index 00000000..4a93f5b8
--- /dev/null
+++ b/adapters/gotaglib/end_to_end_test.go
@@ -0,0 +1,274 @@
+package gotaglib
+
+import (
+	"io/fs"
+	"os"
+	"time"
+
+	"github.com/djherbis/times"
+	"github.com/navidrome/navidrome/model"
+	"github.com/navidrome/navidrome/model/metadata"
+	"github.com/navidrome/navidrome/utils/gg"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+type testFileInfo struct {
+	fs.FileInfo
+}
+
+func (t testFileInfo) BirthTime() time.Time {
+	if ts := times.Get(t.FileInfo); ts.HasBirthTime() {
+		return ts.BirthTime()
+	}
+	return t.FileInfo.ModTime()
+}
+
+var _ = Describe("Extractor", func() {
+	toP := func(name, sortName, mbid string) model.Participant {
+		return model.Participant{
+			Artist: model.Artist{Name: name, SortArtistName: sortName, MbzArtistID: mbid},
+		}
+	}
+
+	roles := []struct {
+		model.Role
+		model.ParticipantList
+	}{
+		{model.RoleComposer, model.ParticipantList{
+			toP("coma a", "a, coma", "bf13b584-f27c-43db-8f42-32898d33d4e2"),
+			toP("comb", "comb", "924039a2-09c6-4d29-9b4f-50cc54447d36"),
+		}},
+		{model.RoleLyricist, model.ParticipantList{
+			toP("la a", "a, la", "c84f648f-68a6-40a2-a0cb-d135b25da3c2"),
+			toP("lb", "lb", "0a7c582d-143a-4540-b4e9-77200835af65"),
+		}},
+		{model.RoleArranger, model.ParticipantList{
+			toP("aa", "", "4605a1d4-8d15-42a3-bd00-9c20e42f71e6"),
+			toP("ab", "", "002f0ff8-77bf-42cc-8216-61a9c43dc145"),
+		}},
+		{model.RoleConductor, model.ParticipantList{
+			toP("cona", "", "af86879b-2141-42af-bad2-389a4dc91489"),
+			toP("conb", "", "3dfa3c70-d7d3-4b97-b953-c298dd305e12"),
+		}},
+		{model.RoleDirector, model.ParticipantList{
+			toP("dia", "", "f943187f-73de-4794-be47-88c66f0fd0f4"),
+			toP("dib", "", "bceb75da-1853-4b3d-b399-b27f0cafc389"),
+		}},
+		{model.RoleEngineer, model.ParticipantList{
+			toP("ea", "", "f634bf6d-d66a-425d-888a-28ad39392759"),
+			toP("eb", "", "243d64ae-d514-44e1-901a-b918d692baee"),
+		}},
+		{model.RoleProducer, model.ParticipantList{
+			toP("pra", "", "d971c8d7-999c-4a5f-ac31-719721ab35d6"),
+			toP("prb", "", "f0a09070-9324-434f-a599-6d25ded87b69"),
+		}},
+		{model.RoleRemixer, model.ParticipantList{
+			toP("ra", "", "c7dc6095-9534-4c72-87cc-aea0103462cf"),
+			toP("rb", "", "8ebeef51-c08c-4736-992f-c37870becedd"),
+		}},
+		{model.RoleDJMixer, model.ParticipantList{
+			toP("dja", "", "d063f13b-7589-4efc-ab7f-c60e6db17247"),
+			toP("djb", "", "3636670c-385f-4212-89c8-0ff51d6bc456"),
+		}},
+		{model.RoleMixer, model.ParticipantList{
+			toP("ma", "", "53fb5a2d-7016-427e-a563-d91819a5f35a"),
+			toP("mb", "", "64c13e65-f0da-4ab9-a300-71ee53b0376a"),
+		}},
+	}
+
+	var e *extractor
+
+	parseTestFile := func(path string) *model.MediaFile {
+		mds, err := e.Parse(path)
+		Expect(err).ToNot(HaveOccurred())
+
+		info, ok := mds[path]
+		Expect(ok).To(BeTrue())
+
+		fileInfo, err := os.Stat(path)
+		Expect(err).ToNot(HaveOccurred())
+		info.FileInfo = testFileInfo{FileInfo: fileInfo}
+
+		metadata := metadata.New(path, info)
+		mf := metadata.ToMediaFile(1, "folderID")
+		return &mf
+	}
+
+	BeforeEach(func() {
+		e = &extractor{fs: os.DirFS(".")}
+	})
+
+	Describe("ReplayGain", func() {
+		DescribeTable("test replaygain end-to-end", func(file string, trackGain, trackPeak, albumGain, albumPeak *float64) {
+			mf := parseTestFile("tests/fixtures/" + file)
+
+			Expect(mf.RGTrackGain).To(Equal(trackGain))
+			Expect(mf.RGTrackPeak).To(Equal(trackPeak))
+			Expect(mf.RGAlbumGain).To(Equal(albumGain))
+			Expect(mf.RGAlbumPeak).To(Equal(albumPeak))
+		},
+			Entry("mp3 with no replaygain", "no_replaygain.mp3", nil, nil, nil, nil),
+			Entry("mp3 with no zero replaygain", "zero_replaygain.mp3", gg.P(0.0), gg.P(1.0), gg.P(0.0), gg.P(1.0)),
+		)
+	})
+
+	Describe("lyrics", func() {
+		makeLyrics := func(code, secondLine string) model.Lyrics {
+			return model.Lyrics{
+				DisplayArtist: "",
+				DisplayTitle:  "",
+				Lang:          code,
+				Line: []model.Line{
+					{Start: gg.P(int64(0)), Value: "This is"},
+					{Start: gg.P(int64(2500)), Value: secondLine},
+				},
+				Offset: nil,
+				Synced: true,
+			}
+		}
+
+		It("should fetch both synced and unsynced lyrics in mixed flac", func() {
+			mf := parseTestFile("tests/fixtures/mixed-lyrics.flac")
+
+			lyrics, err := mf.StructuredLyrics()
+			Expect(err).ToNot(HaveOccurred())
+			Expect(lyrics).To(HaveLen(2))
+
+			Expect(lyrics[0].Synced).To(BeTrue())
+			Expect(lyrics[1].Synced).To(BeFalse())
+		})
+
+		It("should handle mp3 with uslt and sylt", func() {
+			mf := parseTestFile("tests/fixtures/test.mp3")
+
+			lyrics, err := mf.StructuredLyrics()
+			Expect(err).ToNot(HaveOccurred())
+			Expect(lyrics).To(HaveLen(4))
+
+			engSylt := makeLyrics("eng", "English SYLT")
+			engUslt := makeLyrics("eng", "English")
+			unsSylt := makeLyrics("xxx", "unspecified SYLT")
+			unsUslt := makeLyrics("xxx", "unspecified")
+
+			Expect(lyrics).To(ConsistOf(engSylt, engUslt, unsSylt, unsUslt))
+		})
+
+		DescribeTable("format-specific lyrics", func(file string, isId3 bool) {
+			mf := parseTestFile("tests/fixtures/" + file)
+
+			lyrics, err := mf.StructuredLyrics()
+			Expect(err).To(Not(HaveOccurred()))
+			Expect(lyrics).To(HaveLen(2))
+
+			unspec := makeLyrics("xxx", "unspecified")
+			eng := makeLyrics("xxx", "English")
+
+			if isId3 {
+				eng.Lang = "eng"
+			}
+
+			Expect(lyrics).To(Or(
+				Equal(model.LyricList{unspec, eng}),
+				Equal(model.LyricList{eng, unspec})))
+		},
+			Entry("flac", "test.flac", false),
+			Entry("m4a", "test.m4a", false),
+			Entry("ogg", "test.ogg", false),
+			Entry("wma", "test.wma", false),
+			Entry("wv", "test.wv", false),
+			Entry("wav", "test.wav", true),
+			Entry("aiff", "test.aiff", true),
+		)
+	})
+
+	Describe("Participants", func() {
+		DescribeTable("test tags consistent across formats", func(format string) {
+			mf := parseTestFile("tests/fixtures/test." + format)
+
+			for _, data := range roles {
+				role := data.Role
+				artists := data.ParticipantList
+
+				actual := mf.Participants[role]
+				Expect(actual).To(HaveLen(len(artists)))
+
+				for i := range artists {
+					actualArtist := actual[i]
+					expectedArtist := artists[i]
+
+					Expect(actualArtist.Name).To(Equal(expectedArtist.Name))
+					Expect(actualArtist.SortArtistName).To(Equal(expectedArtist.SortArtistName))
+					Expect(actualArtist.MbzArtistID).To(Equal(expectedArtist.MbzArtistID))
+				}
+			}
+
+			if format != "m4a" {
+				performers := mf.Participants[model.RolePerformer]
+				Expect(performers).To(HaveLen(8))
+
+				rules := map[string][]string{
+					"pgaa": {"2fd0b311-9fa8-4ff9-be5d-f6f3d16b835e", "Guitar"},
+					"pgbb": {"223d030b-bf97-4c2a-ad26-b7f7bbe25c93", "Guitar", ""},
+					"pvaa": {"cb195f72-448f-41c8-b962-3f3c13d09d38", "Vocals"},
+					"pvbb": {"60a1f832-8ca2-49f6-8660-84d57f07b520", "Vocals", "Flute"},
+					"pfaa": {"51fb40c-0305-4bf9-a11b-2ee615277725", "", "Flute"},
+				}
+
+				for name, rule := range rules {
+					mbid := rule[0]
+					for i := 1; i < len(rule); i++ {
+						found := false
+
+						for _, mapped := range performers {
+							if mapped.Name == name && mapped.MbzArtistID == mbid && mapped.SubRole == rule[i] {
+								found = true
+								break
+							}
+						}
+
+						Expect(found).To(BeTrue(), "Could not find matching artist")
+					}
+				}
+			}
+		},
+			Entry("FLAC format", "flac"),
+			Entry("M4a format", "m4a"),
+			Entry("OGG format", "ogg"),
+			Entry("WV format", "wv"),
+
+			Entry("MP3 format", "mp3"),
+			Entry("WAV format", "wav"),
+			Entry("AIFF format", "aiff"),
+		)
+
+		It("should parse wma", func() {
+			mf := parseTestFile("tests/fixtures/test.wma")
+
+			for _, data := range roles {
+				role := data.Role
+				artists := data.ParticipantList
+				actual := mf.Participants[role]
+
+				// WMA has no Arranger role
+				if role == model.RoleArranger {
+					Expect(actual).To(HaveLen(0))
+					continue
+				}
+
+				Expect(actual).To(HaveLen(len(artists)), role.String())
+
+				// For some bizarre reason, the order is inverted. We also don't get
+				// sort names or MBIDs
+				for i := range artists {
+					idx := len(artists) - 1 - i
+
+					actualArtist := actual[i]
+					expectedArtist := artists[idx]
+
+					Expect(actualArtist.Name).To(Equal(expectedArtist.Name))
+				}
+			}
+		})
+	})
+})
diff --git a/adapters/gotaglib/gotaglib.go b/adapters/gotaglib/gotaglib.go
new file mode 100644
index 00000000..46c5fa0e
--- /dev/null
+++ b/adapters/gotaglib/gotaglib.go
@@ -0,0 +1,263 @@
+// Package gotaglib provides an alternative metadata extractor using go-taglib,
+// a pure Go (WASM-based) implementation of TagLib.
+//
+// This extractor aims for parity with the CGO-based taglib extractor. It uses
+// TagLib's PropertyMap interface for standard tags. The File handle API provides
+// efficient access to format-specific tags (ID3v2 frames, MP4 atoms, ASF attributes)
+// through a single file open operation.
+//
+// This extractor is registered under the name "gotaglib". It only works with a filesystem
+// (fs.FS) and does not support direct local file paths. Files returned by the filesystem
+// must implement io.ReadSeeker for go-taglib to read them.
+package gotaglib
+
+import (
+	"errors"
+	"io"
+	"io/fs"
+	"strings"
+	"time"
+
+	"github.com/navidrome/navidrome/core/storage/local"
+	"github.com/navidrome/navidrome/model/metadata"
+	"go.senan.xyz/taglib"
+)
+
+type extractor struct {
+	fs fs.FS
+}
+
+func (e extractor) Parse(files ...string) (map[string]metadata.Info, error) {
+	results := make(map[string]metadata.Info)
+	for _, path := range files {
+		props, err := e.extractMetadata(path)
+		if err != nil {
+			continue
+		}
+		results[path] = *props
+	}
+	return results, nil
+}
+
+func (e extractor) Version() string {
+	return "go-taglib (TagLib 2.1.1 WASM)"
+}
+
+func (e extractor) extractMetadata(filePath string) (*metadata.Info, error) {
+	f, close, err := e.openFile(filePath)
+	if err != nil {
+		return nil, err
+	}
+	defer close()
+
+	// Get all tags and properties in one go
+	allTags := f.AllTags()
+	props := f.Properties()
+
+	// Map properties to AudioProperties
+	ap := metadata.AudioProperties{
+		Duration:   props.Length.Round(time.Millisecond * 10),
+		BitRate:    int(props.Bitrate),
+		Channels:   int(props.Channels),
+		SampleRate: int(props.SampleRate),
+		BitDepth:   int(props.BitsPerSample),
+	}
+
+	// Convert normalized tags to lowercase keys (go-taglib returns UPPERCASE keys)
+	normalizedTags := make(map[string][]string, len(allTags.Tags))
+	for key, values := range allTags.Tags {
+		lowerKey := strings.ToLower(key)
+		normalizedTags[lowerKey] = values
+	}
+
+	// Process format-specific raw tags
+	processRawTags(allTags, normalizedTags)
+
+	// Parse track/disc totals from "N/Total" format
+	parseTuple(normalizedTags, "track")
+	parseTuple(normalizedTags, "disc")
+
+	// Adjust some ID3 tags
+	parseLyrics(normalizedTags)
+	parseTIPL(normalizedTags)
+	delete(normalizedTags, "tmcl") // TMCL is already parsed by TagLib
+
+	// Determine if file has embedded picture
+	hasPicture := len(props.Images) > 0
+
+	return &metadata.Info{
+		Tags:            normalizedTags,
+		AudioProperties: ap,
+		HasPicture:      hasPicture,
+	}, nil
+}
+
+// openFile opens the file at filePath using the extractor's filesystem.
+// It returns a TagLib File handle and a cleanup function to close resources.
+func (e extractor) openFile(filePath string) (*taglib.File, func(), error) {
+	// Open the file from the filesystem
+	file, err := e.fs.Open(filePath)
+	if err != nil {
+		return nil, nil, err
+	}
+	rs, isSeekable := file.(io.ReadSeeker)
+	if !isSeekable {
+		file.Close()
+		return nil, nil, errors.New("file is not seekable")
+	}
+	f, err := taglib.OpenStream(rs)
+	if err != nil {
+		file.Close()
+		return nil, nil, err
+	}
+	closeFunc := func() {
+		f.Close()
+		file.Close()
+	}
+	return f, closeFunc, nil
+}
+
+// parseTuple parses track/disc numbers in "N/Total" format and separates them.
+// For example, tracknumber="2/10" becomes tracknumber="2" and tracktotal="10".
+func parseTuple(tags map[string][]string, prop string) {
+	tagName := prop + "number"
+	tagTotal := prop + "total"
+	if value, ok := tags[tagName]; ok && len(value) > 0 {
+		parts := strings.Split(value[0], "/")
+		tags[tagName] = []string{parts[0]}
+		if len(parts) == 2 {
+			tags[tagTotal] = []string{parts[1]}
+		}
+	}
+}
+
+// parseLyrics ensures lyrics tags have a language code.
+// If lyrics exist without a language code, they are moved to "lyrics:xxx".
+func parseLyrics(tags map[string][]string) {
+	lyrics := tags["lyrics"]
+	if len(lyrics) > 0 {
+		tags["lyrics:xxx"] = lyrics
+		delete(tags, "lyrics")
+	}
+}
+
+// processRawTags processes format-specific raw tags based on the detected file format.
+// This handles ID3v2 frames (MP3/WAV/AIFF), MP4 atoms, and ASF attributes.
+func processRawTags(allTags taglib.AllTags, normalizedTags map[string][]string) {
+	switch allTags.Format {
+	case taglib.FormatMPEG, taglib.FormatWAV, taglib.FormatAIFF:
+		parseID3v2Frames(allTags.Raw, normalizedTags)
+	case taglib.FormatMP4:
+		parseMP4Atoms(allTags.Raw, normalizedTags)
+	case taglib.FormatASF:
+		parseASFAttributes(allTags.Raw, normalizedTags)
+	}
+}
+
+// parseID3v2Frames processes ID3v2 raw frames to extract USLT/SYLT with language codes.
+// This extracts language-specific lyrics that the standard Tags() doesn't provide.
+func parseID3v2Frames(rawFrames map[string][]string, tags map[string][]string) {
+	// Process frames that have language-specific data
+	for key, values := range rawFrames {
+		lowerKey := strings.ToLower(key)
+
+		// Handle USLT:xxx and SYLT:xxx (lyrics with language codes)
+		if strings.HasPrefix(lowerKey, "uslt:") || strings.HasPrefix(lowerKey, "sylt:") {
+			parts := strings.SplitN(lowerKey, ":", 2)
+			if len(parts) == 2 && parts[1] != "" {
+				lang := parts[1]
+				lyricsKey := "lyrics:" + lang
+				tags[lyricsKey] = append(tags[lyricsKey], values...)
+			}
+		}
+	}
+
+	// If we found any language-specific lyrics from ID3v2 frames, remove the generic lyrics
+	for key := range tags {
+		if strings.HasPrefix(key, "lyrics:") && key != "lyrics" {
+			delete(tags, "lyrics")
+			break
+		}
+	}
+}
+
+const iTunesKeyPrefix = "----:com.apple.iTunes:"
+
+// parseMP4Atoms processes MP4 raw atoms to get iTunes-specific tags.
+func parseMP4Atoms(rawAtoms map[string][]string, tags map[string][]string) {
+	// Process all atoms and add them to tags
+	for key, values := range rawAtoms {
+		// Strip iTunes prefix and convert to lowercase
+		normalizedKey := strings.TrimPrefix(key, iTunesKeyPrefix)
+		normalizedKey = strings.ToLower(normalizedKey)
+
+		// Only add if the tag doesn't already exist (avoid duplication with PropertyMap)
+		if _, exists := tags[normalizedKey]; !exists {
+			tags[normalizedKey] = values
+		}
+	}
+}
+
+// parseASFAttributes processes ASF raw attributes to get WMA-specific tags.
+func parseASFAttributes(rawAttrs map[string][]string, tags map[string][]string) {
+	// Process all attributes and add them to tags
+	for key, values := range rawAttrs {
+		normalizedKey := strings.ToLower(key)
+
+		// Only add if the tag doesn't already exist (avoid duplication with PropertyMap)
+		if _, exists := tags[normalizedKey]; !exists {
+			tags[normalizedKey] = values
+		}
+	}
+}
+
+// These are the only roles we support, based on Picard's tag map:
+// https://picard-docs.musicbrainz.org/downloads/MusicBrainz_Picard_Tag_Map.html
+var tiplMapping = map[string]string{
+	"arranger": "arranger",
+	"engineer": "engineer",
+	"producer": "producer",
+	"mix":      "mixer",
+	"DJ-mix":   "djmixer",
+}
+
+// parseTIPL parses the ID3v2.4 TIPL frame string, which is received from TagLib in the format:
+//
+//	"arranger Andrew Powell engineer Chris Blair engineer Pat Stapley producer Eric Woolfson".
+//
+// and breaks it down into a map of roles and names, e.g.:
+//
+//	{"arranger": ["Andrew Powell"], "engineer": ["Chris Blair", "Pat Stapley"], "producer": ["Eric Woolfson"]}.
+func parseTIPL(tags map[string][]string) {
+	tipl := tags["tipl"]
+	if len(tipl) == 0 {
+		return
+	}
+	addRole := func(currentRole string, currentValue []string) {
+		if currentRole != "" && len(currentValue) > 0 {
+			role := tiplMapping[currentRole]
+			tags[role] = append(tags[role], strings.Join(currentValue, " "))
+		}
+	}
+	var currentRole string
+	var currentValue []string
+	for _, part := range strings.Split(tipl[0], " ") {
+		if _, ok := tiplMapping[part]; ok {
+			addRole(currentRole, currentValue)
+			currentRole = part
+			currentValue = nil
+			continue
+		}
+		currentValue = append(currentValue, part)
+	}
+	addRole(currentRole, currentValue)
+	delete(tags, "tipl")
+}
+
+var _ local.Extractor = (*extractor)(nil)
+
+func init() {
+	local.RegisterExtractor("taglib", func(fsys fs.FS, baseDir string) local.Extractor {
+		return &extractor{fsys}
+	})
+}
diff --git a/adapters/gotaglib/gotaglib_suite_test.go b/adapters/gotaglib/gotaglib_suite_test.go
new file mode 100644
index 00000000..cc7ddc47
--- /dev/null
+++ b/adapters/gotaglib/gotaglib_suite_test.go
@@ -0,0 +1,17 @@
+package gotaglib
+
+import (
+	"testing"
+
+	"github.com/navidrome/navidrome/log"
+	"github.com/navidrome/navidrome/tests"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+func TestGoTagLib(t *testing.T) {
+	tests.Init(t, true)
+	log.SetLevel(log.LevelFatal)
+	RegisterFailHandler(Fail)
+	RunSpecs(t, "GoTagLib Suite")
+}
diff --git a/adapters/gotaglib/gotaglib_test.go b/adapters/gotaglib/gotaglib_test.go
new file mode 100644
index 00000000..529a8110
--- /dev/null
+++ b/adapters/gotaglib/gotaglib_test.go
@@ -0,0 +1,302 @@
+package gotaglib
+
+import (
+	"io/fs"
+	"os"
+	"strings"
+
+	"github.com/navidrome/navidrome/utils"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("Extractor", func() {
+	var e *extractor
+
+	BeforeEach(func() {
+		e = &extractor{fs: os.DirFS(".")}
+	})
+
+	Describe("Parse", func() {
+		It("correctly parses metadata from all files in folder", func() {
+			mds, err := e.Parse(
+				"tests/fixtures/test.mp3",
+				"tests/fixtures/test.ogg",
+			)
+			Expect(err).NotTo(HaveOccurred())
+			Expect(mds).To(HaveLen(2))
+
+			// Test MP3
+			m := mds["tests/fixtures/test.mp3"]
+			Expect(m.Tags).To(HaveKeyWithValue("title", []string{"Song"}))
+			Expect(m.Tags).To(HaveKeyWithValue("album", []string{"Album"}))
+			Expect(m.Tags).To(HaveKeyWithValue("artist", []string{"Artist"}))
+			Expect(m.Tags).To(HaveKeyWithValue("albumartist", []string{"Album Artist"}))
+
+			Expect(m.HasPicture).To(BeTrue())
+			Expect(m.AudioProperties.Duration.String()).To(Equal("1.02s"))
+			Expect(m.AudioProperties.BitRate).To(Equal(192))
+			Expect(m.AudioProperties.Channels).To(Equal(2))
+			Expect(m.AudioProperties.SampleRate).To(Equal(44100))
+
+			Expect(m.Tags).To(Or(
+				HaveKeyWithValue("compilation", []string{"1"}),
+				HaveKeyWithValue("tcmp", []string{"1"})),
+			)
+			Expect(m.Tags).To(HaveKeyWithValue("genre", []string{"Rock"}))
+			Expect(m.Tags).To(HaveKeyWithValue("date", []string{"2014-05-21"}))
+			Expect(m.Tags).To(HaveKeyWithValue("originaldate", []string{"1996-11-21"}))
+			Expect(m.Tags).To(HaveKeyWithValue("releasedate", []string{"2020-12-31"}))
+			Expect(m.Tags).To(HaveKeyWithValue("discnumber", []string{"1"}))
+			Expect(m.Tags).To(HaveKeyWithValue("disctotal", []string{"2"}))
+			Expect(m.Tags).To(HaveKeyWithValue("comment", []string{"Comment1\nComment2"}))
+			Expect(m.Tags).To(HaveKeyWithValue("bpm", []string{"123"}))
+			Expect(m.Tags).To(HaveKeyWithValue("replaygain_album_gain", []string{"+3.21518 dB"}))
+			Expect(m.Tags).To(HaveKeyWithValue("replaygain_album_peak", []string{"0.9125"}))
+			Expect(m.Tags).To(HaveKeyWithValue("replaygain_track_gain", []string{"-1.48 dB"}))
+			Expect(m.Tags).To(HaveKeyWithValue("replaygain_track_peak", []string{"0.4512"}))
+
+			Expect(m.Tags).To(HaveKeyWithValue("tracknumber", []string{"2"}))
+			Expect(m.Tags).To(HaveKeyWithValue("tracktotal", []string{"10"}))
+
+			Expect(m.Tags).ToNot(HaveKey("lyrics"))
+			Expect(m.Tags).To(Or(HaveKeyWithValue("lyrics:eng", []string{
+				"[00:00.00]This is\n[00:02.50]English SYLT\n",
+				"[00:00.00]This is\n[00:02.50]English",
+			}), HaveKeyWithValue("lyrics:eng", []string{
+				"[00:00.00]This is\n[00:02.50]English",
+				"[00:00.00]This is\n[00:02.50]English SYLT\n",
+			})))
+			Expect(m.Tags).To(Or(HaveKeyWithValue("lyrics:xxx", []string{
+				"[00:00.00]This is\n[00:02.50]unspecified SYLT\n",
+				"[00:00.00]This is\n[00:02.50]unspecified",
+			}), HaveKeyWithValue("lyrics:xxx", []string{
+				"[00:00.00]This is\n[00:02.50]unspecified",
+				"[00:00.00]This is\n[00:02.50]unspecified SYLT\n",
+			})))
+
+			// Test OGG
+			m = mds["tests/fixtures/test.ogg"]
+			Expect(err).To(BeNil())
+			Expect(m.Tags).To(HaveKeyWithValue("fbpm", []string{"141.7"}))
+
+			// TagLib 1.12 returns 18, previous versions return 39.
+			// See https://github.com/taglib/taglib/commit/2f238921824741b2cfe6fbfbfc9701d9827ab06b
+			Expect(m.AudioProperties.BitRate).To(BeElementOf(18, 19, 39, 40, 43, 49))
+			Expect(m.AudioProperties.Channels).To(BeElementOf(2))
+			Expect(m.AudioProperties.SampleRate).To(BeElementOf(8000))
+			Expect(m.HasPicture).To(BeTrue())
+		})
+
+		DescribeTable("Format-Specific tests",
+			func(file, duration string, channels, samplerate, bitdepth int, albumGain, albumPeak, trackGain, trackPeak string, id3Lyrics bool, image bool) {
+				file = "tests/fixtures/" + file
+				mds, err := e.Parse(file)
+				Expect(err).NotTo(HaveOccurred())
+				Expect(mds).To(HaveLen(1))
+
+				m := mds[file]
+
+				Expect(m.HasPicture).To(Equal(image))
+				Expect(m.AudioProperties.Duration.String()).To(Equal(duration))
+				Expect(m.AudioProperties.Channels).To(Equal(channels))
+				Expect(m.AudioProperties.SampleRate).To(Equal(samplerate))
+				Expect(m.AudioProperties.BitDepth).To(Equal(bitdepth))
+
+				Expect(m.Tags).To(Or(
+					HaveKeyWithValue("replaygain_album_gain", []string{albumGain}),
+					HaveKeyWithValue("----:com.apple.itunes:replaygain_album_gain", []string{albumGain}),
+				))
+
+				Expect(m.Tags).To(Or(
+					HaveKeyWithValue("replaygain_album_peak", []string{albumPeak}),
+					HaveKeyWithValue("----:com.apple.itunes:replaygain_album_peak", []string{albumPeak}),
+				))
+				Expect(m.Tags).To(Or(
+					HaveKeyWithValue("replaygain_track_gain", []string{trackGain}),
+					HaveKeyWithValue("----:com.apple.itunes:replaygain_track_gain", []string{trackGain}),
+				))
+				Expect(m.Tags).To(Or(
+					HaveKeyWithValue("replaygain_track_peak", []string{trackPeak}),
+					HaveKeyWithValue("----:com.apple.itunes:replaygain_track_peak", []string{trackPeak}),
+				))
+
+				Expect(m.Tags).To(HaveKeyWithValue("title", []string{"Title"}))
+				Expect(m.Tags).To(HaveKeyWithValue("album", []string{"Album"}))
+				Expect(m.Tags).To(HaveKeyWithValue("artist", []string{"Artist"}))
+				Expect(m.Tags).To(HaveKeyWithValue("albumartist", []string{"Album Artist"}))
+				Expect(m.Tags).To(HaveKeyWithValue("genre", []string{"Rock"}))
+				Expect(m.Tags).To(HaveKeyWithValue("date", []string{"2014"}))
+
+				Expect(m.Tags).To(HaveKeyWithValue("bpm", []string{"123"}))
+				Expect(m.Tags).To(Or(
+					HaveKeyWithValue("tracknumber", []string{"3"}),
+					HaveKeyWithValue("tracknumber", []string{"3/10"}),
+				))
+				if !strings.HasSuffix(file, "test.wma") {
+					// TODO Not sure why this is not working for WMA
+					Expect(m.Tags).To(HaveKeyWithValue("tracktotal", []string{"10"}))
+				}
+				Expect(m.Tags).To(Or(
+					HaveKeyWithValue("discnumber", []string{"1"}),
+					HaveKeyWithValue("discnumber", []string{"1/2"}),
+				))
+				Expect(m.Tags).To(HaveKeyWithValue("disctotal", []string{"2"}))
+
+				// WMA does not have a "compilation" tag, but "wm/iscompilation"
+				Expect(m.Tags).To(Or(
+					HaveKeyWithValue("compilation", []string{"1"}),
+					HaveKeyWithValue("wm/iscompilation", []string{"1"})),
+				)
+
+				if id3Lyrics {
+					Expect(m.Tags).To(HaveKeyWithValue("lyrics:eng", []string{
+						"[00:00.00]This is\n[00:02.50]English",
+					}))
+					Expect(m.Tags).To(HaveKeyWithValue("lyrics:xxx", []string{
+						"[00:00.00]This is\n[00:02.50]unspecified",
+					}))
+				} else {
+					Expect(m.Tags).To(HaveKeyWithValue("lyrics:xxx", []string{
+						"[00:00.00]This is\n[00:02.50]unspecified",
+						"[00:00.00]This is\n[00:02.50]English",
+					}))
+				}
+
+				Expect(m.Tags).To(HaveKeyWithValue("comment", []string{"Comment1\nComment2"}))
+			},
+
+			// ffmpeg -f lavfi -i "sine=frequency=1200:duration=1" test.flac
+			Entry("correctly parses flac tags", "test.flac", "1s", 1, 44100, 16, "+4.06 dB", "0.12496948", "+4.06 dB", "0.12496948", false, true),
+
+			Entry("correctly parses m4a (aac) gain tags", "01 Invisible (RED) Edit Version.m4a", "1.04s", 2, 44100, 16, "0.37", "0.48", "0.37", "0.48", false, true),
+			Entry("correctly parses m4a (aac) gain tags (uppercase)", "test.m4a", "1.04s", 2, 44100, 16, "0.37", "0.48", "0.37", "0.48", false, true),
+			Entry("correctly parses ogg (vorbis) tags", "test.ogg", "1.04s", 2, 8000, 0, "+7.64 dB", "0.11772506", "+7.64 dB", "0.11772506", false, true),
+
+			// ffmpeg -f lavfi -i "sine=frequency=900:duration=1" test.wma
+			// Weird note: for the tag parsing to work, the lyrics are actually stored in the reverse order
+			Entry("correctly parses wma/asf tags", "test.wma", "1.02s", 1, 44100, 16, "3.27 dB", "0.132914", "3.27 dB", "0.132914", false, true),
+
+			// ffmpeg -f lavfi -i "sine=frequency=800:duration=1" test.wv
+			Entry("correctly parses wv (wavpak) tags", "test.wv", "1s", 1, 44100, 16, "3.43 dB", "0.125061", "3.43 dB", "0.125061", false, true),
+
+			// ffmpeg -f lavfi -i "sine=frequency=1000:duration=1" test.wav
+			Entry("correctly parses wav tags", "test.wav", "1s", 1, 44100, 16, "3.06 dB", "0.125056", "3.06 dB", "0.125056", true, true),
+
+			// ffmpeg -f lavfi -i "sine=frequency=1400:duration=1" test.aiff
+			Entry("correctly parses aiff tags", "test.aiff", "1s", 1, 44100, 16, "2.00 dB", "0.124972", "2.00 dB", "0.124972", true, true),
+		)
+
+		// Skip these tests when running as root
+		Context("Access Forbidden", func() {
+			var accessForbiddenFile string
+			var RegularUserContext = XContext
+			var isRegularUser = os.Getuid() != 0
+			if isRegularUser {
+				RegularUserContext = Context
+			}
+
+			// Only run permission tests if we are not root
+			RegularUserContext("when run without root privileges", func() {
+				BeforeEach(func() {
+					// Use root fs for absolute paths in temp directory
+					e = &extractor{fs: os.DirFS("/")}
+					accessForbiddenFile = utils.TempFileName("access_forbidden-", ".mp3")
+
+					f, err := os.OpenFile(accessForbiddenFile, os.O_WRONLY|os.O_CREATE, 0222)
+					Expect(err).ToNot(HaveOccurred())
+
+					DeferCleanup(func() {
+						Expect(f.Close()).To(Succeed())
+						Expect(os.Remove(accessForbiddenFile)).To(Succeed())
+					})
+				})
+
+				It("correctly handle unreadable file due to insufficient read permission", func() {
+					// Strip leading slash for DirFS rooted at "/"
+					_, err := e.extractMetadata(accessForbiddenFile[1:])
+					Expect(err).To(MatchError(os.ErrPermission))
+				})
+
+				It("skips the file if it cannot be read", func() {
+					// Get current working directory to construct paths relative to root
+					cwd, err := os.Getwd()
+					Expect(err).ToNot(HaveOccurred())
+					// Strip leading slash for DirFS rooted at "/"
+					files := []string{
+						cwd[1:] + "/tests/fixtures/test.mp3",
+						cwd[1:] + "/tests/fixtures/test.ogg",
+						accessForbiddenFile[1:],
+					}
+					mds, err := e.Parse(files...)
+					Expect(err).NotTo(HaveOccurred())
+					Expect(mds).To(HaveLen(2))
+					Expect(mds).ToNot(HaveKey(accessForbiddenFile[1:]))
+				})
+			})
+		})
+
+	})
+
+	Describe("Error Checking", func() {
+		It("returns a generic ErrPath if file does not exist", func() {
+			testFilePath := "tests/fixtures/NON_EXISTENT.ogg"
+			_, err := e.extractMetadata(testFilePath)
+			Expect(err).To(MatchError(fs.ErrNotExist))
+		})
+		It("does not throw a SIGSEGV error when reading a file with an invalid frame", func() {
+			// File has an empty TDAT frame
+			md, err := e.extractMetadata("tests/fixtures/invalid-files/test-invalid-frame.mp3")
+			Expect(err).ToNot(HaveOccurred())
+			Expect(md.Tags).To(HaveKeyWithValue("albumartist", []string{"Elvis Presley"}))
+		})
+	})
+
+	Describe("parseTIPL", func() {
+		var tags map[string][]string
+
+		BeforeEach(func() {
+			tags = make(map[string][]string)
+		})
+
+		Context("when the TIPL string is populated", func() {
+			It("correctly parses roles and names", func() {
+				tags["tipl"] = []string{"arranger Andrew Powell DJ-mix François Kevorkian DJ-mix Jane Doe engineer Chris Blair"}
+				parseTIPL(tags)
+				Expect(tags["arranger"]).To(ConsistOf("Andrew Powell"))
+				Expect(tags["engineer"]).To(ConsistOf("Chris Blair"))
+				Expect(tags["djmixer"]).To(ConsistOf("François Kevorkian", "Jane Doe"))
+			})
+
+			It("handles multiple names for a single role", func() {
+				tags["tipl"] = []string{"engineer Pat Stapley producer Eric Woolfson engineer Chris Blair"}
+				parseTIPL(tags)
+				Expect(tags["producer"]).To(ConsistOf("Eric Woolfson"))
+				Expect(tags["engineer"]).To(ConsistOf("Pat Stapley", "Chris Blair"))
+			})
+
+			It("discards roles without names", func() {
+				tags["tipl"] = []string{"engineer Pat Stapley producer engineer Chris Blair"}
+				parseTIPL(tags)
+				Expect(tags).ToNot(HaveKey("producer"))
+				Expect(tags["engineer"]).To(ConsistOf("Pat Stapley", "Chris Blair"))
+			})
+		})
+
+		Context("when the TIPL string is empty", func() {
+			It("does nothing", func() {
+				tags["tipl"] = []string{""}
+				parseTIPL(tags)
+				Expect(tags).To(BeEmpty())
+			})
+		})
+
+		Context("when the TIPL is not present", func() {
+			It("does nothing", func() {
+				parseTIPL(tags)
+				Expect(tags).To(BeEmpty())
+			})
+		})
+	})
+
+})
diff --git a/adapters/taglib/end_to_end_test.go b/adapters/taglib/end_to_end_test.go
index e4d94bb2..265f258f 100644
--- a/adapters/taglib/end_to_end_test.go
+++ b/adapters/taglib/end_to_end_test.go
@@ -151,11 +151,7 @@ var _ = Describe("Extractor", func() {
 			unsSylt := makeLyrics("xxx", "unspecified SYLT")
 			unsUslt := makeLyrics("xxx", "unspecified")
 
-			// Why is the order inconsistent between runs? Nobody knows
-			Expect(lyrics).To(Or(
-				Equal(model.LyricList{engSylt, engUslt, unsSylt, unsUslt}),
-				Equal(model.LyricList{unsSylt, unsUslt, engSylt, engUslt}),
-			))
+			Expect(lyrics).To(ConsistOf(engSylt, engUslt, unsSylt, unsUslt))
 		})
 
 		DescribeTable("format-specific lyrics", func(file string, isId3 bool) {
diff --git a/adapters/taglib/taglib.go b/adapters/taglib/taglib.go
index d32adf4e..ac299ea2 100644
--- a/adapters/taglib/taglib.go
+++ b/adapters/taglib/taglib.go
@@ -168,7 +168,7 @@ func parseTIPL(tags map[string][]string) {
 var _ local.Extractor = (*extractor)(nil)
 
 func init() {
-	local.RegisterExtractor("taglib", func(_ fs.FS, baseDir string) local.Extractor {
+	local.RegisterExtractor("legacy-taglib", func(_ fs.FS, baseDir string) local.Extractor {
 		// ignores fs, as taglib extractor only works with local files
 		return &extractor{baseDir}
 	})
diff --git a/adapters/taglib/taglib_test.go b/adapters/taglib/taglib_test.go
index f24c0e83..f524f77e 100644
--- a/adapters/taglib/taglib_test.go
+++ b/adapters/taglib/taglib_test.go
@@ -80,12 +80,11 @@ var _ = Describe("Extractor", func() {
 			Expect(err).To(BeNil())
 			Expect(m.Tags).To(HaveKeyWithValue("fbpm", []string{"141.7"}))
 
-			// TabLib 1.12 returns 18, previous versions return 39.
+			// TagLib 1.12 returns 18, previous versions return 39.
 			// See https://github.com/taglib/taglib/commit/2f238921824741b2cfe6fbfbfc9701d9827ab06b
 			Expect(m.AudioProperties.BitRate).To(BeElementOf(18, 19, 39, 40, 43, 49))
 			Expect(m.AudioProperties.Channels).To(BeElementOf(2))
 			Expect(m.AudioProperties.SampleRate).To(BeElementOf(8000))
-			Expect(m.AudioProperties.SampleRate).To(BeElementOf(8000))
 			Expect(m.HasPicture).To(BeTrue())
 		})
 
@@ -106,7 +105,7 @@ var _ = Describe("Extractor", func() {
 
 				Expect(m.Tags).To(Or(
 					HaveKeyWithValue("replaygain_album_gain", []string{albumGain}),
-					HaveKeyWithValue("----:com.apple.itunes:replaygain_track_gain", []string{albumGain}),
+					HaveKeyWithValue("----:com.apple.itunes:replaygain_album_gain", []string{albumGain}),
 				))
 
 				Expect(m.Tags).To(Or(
diff --git a/cmd/root.go b/cmd/root.go
index f5f7ff0c..74a15abc 100644
--- a/cmd/root.go
+++ b/cmd/root.go
@@ -24,6 +24,7 @@ import (
 
 	// Import adapters to register them
 	_ "github.com/navidrome/navidrome/adapters/deezer"
+	_ "github.com/navidrome/navidrome/adapters/gotaglib"
 	_ "github.com/navidrome/navidrome/adapters/lastfm"
 	_ "github.com/navidrome/navidrome/adapters/listenbrainz"
 	_ "github.com/navidrome/navidrome/adapters/spotify"
diff --git a/cmd/wire_gen.go b/cmd/wire_gen.go
index a016d7bd..7a9d38d9 100644
--- a/cmd/wire_gen.go
+++ b/cmd/wire_gen.go
@@ -33,6 +33,7 @@ import (
 
 import (
 	_ "github.com/navidrome/navidrome/adapters/deezer"
+	_ "github.com/navidrome/navidrome/adapters/gotaglib"
 	_ "github.com/navidrome/navidrome/adapters/lastfm"
 	_ "github.com/navidrome/navidrome/adapters/listenbrainz"
 	_ "github.com/navidrome/navidrome/adapters/spotify"
diff --git a/conf/configuration.go b/conf/configuration.go
index 6822a1a9..29e6582e 100644
--- a/conf/configuration.go
+++ b/conf/configuration.go
@@ -366,10 +366,6 @@ func Load(noConfigDump bool) {
 		disableExternalServices()
 	}
 
-	if Server.Scanner.Extractor != consts.DefaultScannerExtractor {
-		log.Warn(fmt.Sprintf("Extractor '%s' is not implemented, using 'taglib'", Server.Scanner.Extractor))
-		Server.Scanner.Extractor = consts.DefaultScannerExtractor
-	}
 	logDeprecatedOptions("Scanner.GenreSeparators", "")
 	logDeprecatedOptions("Scanner.GroupAlbumReleases", "")
 	logDeprecatedOptions("DevEnableBufferedScrobble", "") // Deprecated: Buffered scrobbling is now always enabled and this option is ignored
diff --git a/core/library_test.go b/core/library_test.go
index 16972da7..175d9c37 100644
--- a/core/library_test.go
+++ b/core/library_test.go
@@ -9,7 +9,7 @@ import (
 	"sync"
 
 	"github.com/deluan/rest"
-	_ "github.com/navidrome/navidrome/adapters/taglib" // Register taglib extractor
+	_ "github.com/navidrome/navidrome/adapters/gotaglib" // Register taglib extractor
 	"github.com/navidrome/navidrome/conf/configtest"
 	"github.com/navidrome/navidrome/core"
 	_ "github.com/navidrome/navidrome/core/storage/local" // Register local storage
diff --git a/go.mod b/go.mod
index eacc4131..8d4ee3f7 100644
--- a/go.mod
+++ b/go.mod
@@ -2,8 +2,13 @@ module github.com/navidrome/navidrome
 
 go 1.25
 
-// Fork to fix https://github.com/navidrome/navidrome/issues/3254
-replace github.com/dhowden/tag v0.0.0-20240417053706-3d75831295e8 => github.com/deluan/tag v0.0.0-20241002021117-dfe5e6ea396d
+replace (
+	// Fork to fix https://github.com/navidrome/navidrome/issues/3254
+	github.com/dhowden/tag v0.0.0-20240417053706-3d75831295e8 => github.com/deluan/tag v0.0.0-20241002021117-dfe5e6ea396d
+
+	// Fork to implement raw tags support
+	go.senan.xyz/taglib => github.com/deluan/go-taglib v0.0.0-20260118171208-db06bab917c7
+)
 
 require (
 	github.com/Masterminds/squirrel v1.5.4
@@ -60,6 +65,7 @@ require (
 	github.com/tetratelabs/wazero v1.11.0
 	github.com/unrolled/secure v1.17.0
 	github.com/xrash/smetrics v0.0.0-20250705151800-55b8f293f342
+	go.senan.xyz/taglib v0.0.0-00010101000000-000000000000
 	go.uber.org/goleak v1.3.0
 	golang.org/x/image v0.35.0
 	golang.org/x/net v0.49.0
diff --git a/go.sum b/go.sum
index 255a1bd0..89aad7a9 100644
--- a/go.sum
+++ b/go.sum
@@ -36,6 +36,8 @@ github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc h1:U9qPSI2PIWSS1
 github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/decred/dcrd/dcrec/secp256k1/v4 v4.4.0 h1:NMZiJj8QnKe1LgsbDayM4UoHwbvwDRwnI3hwNaAHRnc=
 github.com/decred/dcrd/dcrec/secp256k1/v4 v4.4.0/go.mod h1:ZXNYxsqcloTdSy/rNShjYzMhyjf0LaoftYK0p+A3h40=
+github.com/deluan/go-taglib v0.0.0-20260118171208-db06bab917c7 h1:ICwI2s4BQdDgp+TY2mAf0jMB7B2hgML7IsSAKTuTRBk=
+github.com/deluan/go-taglib v0.0.0-20260118171208-db06bab917c7/go.mod h1:sKDN0U4qXDlq6LFK+aOAkDH4Me5nDV1V/A4B+B69xBA=
 github.com/deluan/rest v0.0.0-20211102003136-6260bc399cbf h1:tb246l2Zmpt/GpF9EcHCKTtwzrd0HGfEmoODFA/qnk4=
 github.com/deluan/rest v0.0.0-20211102003136-6260bc399cbf/go.mod h1:tSgDythFsl0QgS/PFWfIZqcJKnkADWneY80jaVRlqK8=
 github.com/deluan/sanitize v0.0.0-20241120162836-fdfd8fdfaa55 h1:wSCnggTs2f2ji6nFwQmfwgINcmSMj0xF0oHnoyRSPe4=