Add Go API for speaker identification (#718)

.github/workflows/test-go-package.yaml

@@ -66,6 +66,12 @@ jobs:
         run: |
           gcc --version
 
+      - name: Test speaker identification
+        shell: bash
+        run: |
+          cd go-api-examples/speaker-identification
+          ./run.sh
+
       - name: Test non-streaming TTS (Linux/macOS)
         if: matrix.os != 'windows-latest'
         shell: bash

.github/workflows/test-go.yaml

@@ -74,6 +74,12 @@ jobs:
           go mod tidy
           go build
 
+      - name: Test speaker identification
+        shell: bash
+        run: |
+          cd scripts/go/_internal/speaker-identification/
+          ./run.sh
+
       - name: Test non-streaming TTS (macOS)
         shell: bash
         run: |

.gitignore

@@ -88,3 +88,5 @@ vits-mms-*
 *.tar.bz2
 sherpa-onnx-paraformer-trilingual-zh-cantonese-en
 sr-data
+*xcworkspace/xcuserdata/*
+

go-api-examples/README.md

@@ -26,4 +26,8 @@ for details.
 - [./vad-spoken-language-identification](./vad-spoken-language-identification) It shows how to use silero VAD + Whisper
   for spoken language identification.
 
+- [./speaker-identification](./speaker-identification) It shows how to use Go API for speaker identification.
+
+- [./vad-speaker-identification](./vad-speaker-identification) It shows how to use Go API for VAD + speaker identification.
+
 [sherpa-onnx]: https://github.com/k2-fsa/sherpa-onnx

go-api-examples/speaker-identification/go.mod

@@ -0,0 +1,3 @@
+module speaker-identification
+
+go 1.12

go-api-examples/speaker-identification/main.go

@@ -0,0 +1,146 @@
+package main
+
+import (
+	sherpa "github.com/k2-fsa/sherpa-onnx-go/sherpa_onnx"
+	"log"
+)
+
+func createSpeakerEmbeddingExtractor() *sherpa.SpeakerEmbeddingExtractor {
+	config := sherpa.SpeakerEmbeddingExtractorConfig{}
+
+	// Please download the model from
+	// https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-recongition-models/3dspeaker_speech_campplus_sv_zh-cn_16k-common.onnx
+	//
+	// You can find more models at
+	// https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-recongition-models
+
+	config.Model = "./3dspeaker_speech_campplus_sv_zh-cn_16k-common.onnx"
+	config.NumThreads = 1
+	config.Debug = 1
+	config.Provider = "cpu"
+
+	ex := sherpa.NewSpeakerEmbeddingExtractor(&config)
+	return ex
+}
+
+func computeEmbeddings(ex *sherpa.SpeakerEmbeddingExtractor, files []string) [][]float32 {
+	embeddings := make([][]float32, len(files))
+
+	for i, f := range files {
+		wave := sherpa.ReadWave(f)
+
+		stream := ex.CreateStream()
+		defer sherpa.DeleteOnlineStream(stream)
+		stream.AcceptWaveform(wave.SampleRate, wave.Samples)
+		stream.InputFinished()
+		embeddings[i] = ex.Compute(stream)
+	}
+
+	return embeddings
+
+}
+
+func registerSpeakers(ex *sherpa.SpeakerEmbeddingExtractor, manager *sherpa.SpeakerEmbeddingManager) {
+	// Please download the test data from
+	// https://github.com/csukuangfj/sr-data
+	spk1_files := []string{
+		"./sr-data/enroll/fangjun-sr-1.wav",
+		"./sr-data/enroll/fangjun-sr-2.wav",
+		"./sr-data/enroll/fangjun-sr-3.wav",
+	}
+
+	spk2_files := []string{
+		"./sr-data/enroll/leijun-sr-1.wav",
+		"./sr-data/enroll/leijun-sr-2.wav",
+	}
+
+	spk1_embeddings := computeEmbeddings(ex, spk1_files)
+	spk2_embeddings := computeEmbeddings(ex, spk2_files)
+
+	ok := manager.RegisterV("fangjun", spk1_embeddings)
+	if !ok {
+		panic("Failed to register fangjun")
+	}
+
+	ok = manager.RegisterV("leijun", spk2_embeddings)
+	if !ok {
+		panic("Failed to register leijun")
+	}
+
+	if !manager.Contains("fangjun") {
+		panic("Failed to find fangjun")
+	}
+
+	if !manager.Contains("leijun") {
+		panic("Failed to find leijun")
+	}
+
+	if manager.NumSpeakers() != 2 {
+		panic("There should be only 2 speakers")
+	}
+
+	all_speakers := manager.AllSpeakers()
+	log.Printf("All speakers: %v\n", all_speakers)
+}
+
+func main() {
+	log.SetFlags(log.LstdFlags | log.Lmicroseconds)
+
+	ex := createSpeakerEmbeddingExtractor()
+	defer sherpa.DeleteSpeakerEmbeddingExtractor(ex)
+
+	manager := sherpa.NewSpeakerEmbeddingManager(ex.Dim())
+	defer sherpa.DeleteSpeakerEmbeddingManager(manager)
+	registerSpeakers(ex, manager)
+
+	// Please download the test data from
+	// https://github.com/csukuangfj/sr-data
+	test1 := "./sr-data/test/fangjun-test-sr-1.wav"
+	embeddings := computeEmbeddings(ex, []string{test1})[0]
+	threshold := float32(0.6)
+	name := manager.Search(embeddings, threshold)
+	if len(name) > 0 {
+		log.Printf("%v matches %v", test1, name)
+	} else {
+		log.Printf("No matches found for %v", test1)
+	}
+
+	test2 := "./sr-data/test/leijun-test-sr-1.wav"
+	embeddings = computeEmbeddings(ex, []string{test2})[0]
+	name = manager.Search(embeddings, threshold)
+	if len(name) > 0 {
+		log.Printf("%v matches %v", test2, name)
+	} else {
+		log.Printf("No matches found for %v", test2)
+	}
+
+	test3 := "./sr-data/test/liudehua-test-sr-1.wav"
+	embeddings = computeEmbeddings(ex, []string{test3})[0]
+	name = manager.Search(embeddings, threshold)
+	if len(name) > 0 {
+		log.Printf("%v matches %v", test3, name)
+	} else {
+		log.Printf("No matches found for %v", test3)
+	}
+
+	if !manager.Remove("fangjun") {
+		panic("Failed to deregister fangjun")
+	} else {
+		log.Print("fangjun deregistered\n")
+	}
+
+	test1 = "./sr-data/test/fangjun-test-sr-1.wav"
+	embeddings = computeEmbeddings(ex, []string{test1})[0]
+	name = manager.Search(embeddings, threshold)
+	if len(name) > 0 {
+		log.Printf("%v matches %v", test1, name)
+	} else {
+		log.Printf("No matches found for %v", test1)
+	}
+}
+
+func chk(err error) {
+	if err != nil {
+		panic(err)
+	}
+}

go-api-examples/speaker-identification/run.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+
+if [ ! -f ./3dspeaker_speech_campplus_sv_zh-cn_16k-common.onnx ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-recongition-models/3dspeaker_speech_campplus_sv_zh-cn_16k-common.onnx
+fi
+
+if [ ! -f ./sr-data/enroll/fangjun-sr-1.wav ]; then
+  git clone https://github.com/csukuangfj/sr-data
+fi
+
+go mod tidy
+go build
+./speaker-identification

go-api-examples/vad-asr-paraformer/main.go

@@ -104,7 +104,7 @@ func main() {
 
 			duration := float32(len(speechSegment.Samples)) / float32(config.SampleRate)
 
-			audio := &sherpa.GeneratedAudio{}
+			audio := &sherpa.Wave{}
 			audio.Samples = speechSegment.Samples
 			audio.SampleRate = config.SampleRate
 
@@ -120,7 +120,7 @@ func main() {
 	chk(s.Stop())
 }
 
-func decode(recognizer *sherpa.OfflineRecognizer, audio *sherpa.GeneratedAudio, id int) {
+func decode(recognizer *sherpa.OfflineRecognizer, audio *sherpa.Wave, id int) {
 	stream := sherpa.NewOfflineStream(recognizer)
 	defer sherpa.DeleteOfflineStream(stream)
 	stream.AcceptWaveform(audio.SampleRate, audio.Samples)

go-api-examples/vad-asr-whisper/main.go

@@ -102,7 +102,7 @@ func main() {
 
 			duration := float32(len(speechSegment.Samples)) / float32(config.SampleRate)
 
-			audio := &sherpa.GeneratedAudio{}
+			audio := &sherpa.Wave{}
 			audio.Samples = speechSegment.Samples
 			audio.SampleRate = config.SampleRate
 
@@ -118,7 +118,7 @@ func main() {
 	chk(s.Stop())
 }
 
-func decode(recognizer *sherpa.OfflineRecognizer, audio *sherpa.GeneratedAudio, id int) {
+func decode(recognizer *sherpa.OfflineRecognizer, audio *sherpa.Wave, id int) {
 	stream := sherpa.NewOfflineStream(recognizer)
 	defer sherpa.DeleteOfflineStream(stream)
 	stream.AcceptWaveform(audio.SampleRate, audio.Samples)

go-api-examples/vad-speaker-identification/go.mod

@@ -0,0 +1,3 @@
+module vad-speaker-identification
+
+go 1.12

go-api-examples/vad-speaker-identification/main.go

@@ -0,0 +1,221 @@
+package main
+
+import (
+	"fmt"
+	"github.com/gordonklaus/portaudio"
+	sherpa "github.com/k2-fsa/sherpa-onnx-go/sherpa_onnx"
+	"log"
+)
+
+func createSpeakerEmbeddingExtractor() *sherpa.SpeakerEmbeddingExtractor {
+	config := sherpa.SpeakerEmbeddingExtractorConfig{}
+
+	// Please download the model from
+	// https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-recongition-models/3dspeaker_speech_campplus_sv_zh-cn_16k-common.onnx
+	//
+	// You can find more models at
+	// https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-recongition-models
+
+	config.Model = "./3dspeaker_speech_campplus_sv_zh-cn_16k-common.onnx"
+	config.NumThreads = 2
+	config.Debug = 1
+	config.Provider = "cpu"
+
+	ex := sherpa.NewSpeakerEmbeddingExtractor(&config)
+	return ex
+}
+
+func computeEmbeddings(ex *sherpa.SpeakerEmbeddingExtractor, files []string) [][]float32 {
+	embeddings := make([][]float32, len(files))
+
+	for i, f := range files {
+		wave := sherpa.ReadWave(f)
+
+		stream := ex.CreateStream()
+		defer sherpa.DeleteOnlineStream(stream)
+		stream.AcceptWaveform(wave.SampleRate, wave.Samples)
+		stream.InputFinished()
+		embeddings[i] = ex.Compute(stream)
+	}
+
+	return embeddings
+
+}
+
+func registerSpeakers(ex *sherpa.SpeakerEmbeddingExtractor, manager *sherpa.SpeakerEmbeddingManager) {
+	// Please download the test data from
+	// https://github.com/csukuangfj/sr-data
+	spk1_files := []string{
+		"./sr-data/enroll/fangjun-sr-1.wav",
+		"./sr-data/enroll/fangjun-sr-2.wav",
+		"./sr-data/enroll/fangjun-sr-3.wav",
+	}
+
+	spk2_files := []string{
+		"./sr-data/enroll/leijun-sr-1.wav",
+		"./sr-data/enroll/leijun-sr-2.wav",
+	}
+
+	spk1_embeddings := computeEmbeddings(ex, spk1_files)
+	spk2_embeddings := computeEmbeddings(ex, spk2_files)
+
+	ok := manager.RegisterV("fangjun", spk1_embeddings)
+	if !ok {
+		panic("Failed to register fangjun")
+	}
+
+	ok = manager.RegisterV("leijun", spk2_embeddings)
+	if !ok {
+		panic("Failed to register leijun")
+	}
+
+	if !manager.Contains("fangjun") {
+		panic("Failed to find fangjun")
+	}
+
+	if !manager.Contains("leijun") {
+		panic("Failed to find leijun")
+	}
+
+	if manager.NumSpeakers() != 2 {
+		panic("There should be only 2 speakers")
+	}
+
+	all_speakers := manager.AllSpeakers()
+	log.Printf("All speakers: %v\n", all_speakers)
+}
+
+func createVad() *sherpa.VoiceActivityDetector {
+	config := sherpa.VadModelConfig{}
+
+	// Please download silero_vad.onnx from
+	// https://github.com/snakers4/silero-vad/blob/master/files/silero_vad.onnx
+
+	config.SileroVad.Model = "./silero_vad.onnx"
+	config.SileroVad.Threshold = 0.5
+	config.SileroVad.MinSilenceDuration = 0.5
+	config.SileroVad.MinSpeechDuration = 0.5
+	config.SileroVad.WindowSize = 512
+	config.SampleRate = 16000
+	config.NumThreads = 1
+	config.Provider = "cpu"
+	config.Debug = 1
+
+	var bufferSizeInSeconds float32 = 20
+
+	vad := sherpa.NewVoiceActivityDetector(&config, bufferSizeInSeconds)
+	return vad
+}
+
+func main() {
+	log.SetFlags(log.LstdFlags | log.Lmicroseconds)
+
+	vad := createVad()
+	defer sherpa.DeleteVoiceActivityDetector(vad)
+
+	ex := createSpeakerEmbeddingExtractor()
+	defer sherpa.DeleteSpeakerEmbeddingExtractor(ex)
+
+	manager := sherpa.NewSpeakerEmbeddingManager(ex.Dim())
+	defer sherpa.DeleteSpeakerEmbeddingManager(manager)
+	registerSpeakers(ex, manager)
+
+	err := portaudio.Initialize()
+	if err != nil {
+		log.Fatalf("Unable to initialize portaudio: %v\n", err)
+	}
+	defer portaudio.Terminate()
+
+	default_device, err := portaudio.DefaultInputDevice()
+	if err != nil {
+		log.Fatal("Failed to get default input device: %v\n", err)
+	}
+	log.Printf("Selected default input device: %s\n", default_device.Name)
+	param := portaudio.StreamParameters{}
+	param.Input.Device = default_device
+	param.Input.Channels = 1
+	param.Input.Latency = default_device.DefaultHighInputLatency
+
+	param.SampleRate = 16000
+	param.FramesPerBuffer = 0
+	param.Flags = portaudio.ClipOff
+
+	// you can choose another value for 0.1 if you want
+	samplesPerCall := int32(param.SampleRate * 0.1) // 0.1 second
+	samples := make([]float32, samplesPerCall)
+
+	s, err := portaudio.OpenStream(param, samples)
+	if err != nil {
+		log.Fatalf("Failed to open the stream")
+	}
+
+	defer s.Close()
+	chk(s.Start())
+
+	log.Print("Started! Please speak")
+	printed := false
+
+	k := 0
+	for {
+		chk(s.Read())
+		vad.AcceptWaveform(samples)
+
+		if vad.IsSpeech() && !printed {
+			printed = true
+			log.Print("Detected speech\n")
+		}
+
+		if !vad.IsSpeech() {
+			printed = false
+		}
+
+		for !vad.IsEmpty() {
+			speechSegment := vad.Front()
+			vad.Pop()
+
+			audio := &sherpa.Wave{}
+			audio.Samples = speechSegment.Samples
+			audio.SampleRate = 16000
+
+			// Now decode it
+			go decode(ex, manager, audio, k)
+
+			k += 1
+		}
+	}
+
+	chk(s.Stop())
+
+}
+
+func chk(err error) {
+	if err != nil {
+		panic(err)
+	}
+}
+
+func decode(ex *sherpa.SpeakerEmbeddingExtractor, manager *sherpa.SpeakerEmbeddingManager, audio *sherpa.GeneratedAudio, id int) {
+	stream := ex.CreateStream()
+	defer sherpa.DeleteOnlineStream(stream)
+
+	stream.AcceptWaveform(audio.SampleRate, audio.Samples)
+	stream.InputFinished()
+	embeddings := ex.Compute(stream)
+	threshold := float32(0.5)
+	name := manager.Search(embeddings, threshold)
+	if len(name) > 0 {
+		log.Printf("Found speaker: %v\n", name)
+	} else {
+		log.Print("Unknown speaker\n")
+		name = "Unknown"
+	}
+
+	duration := float32(len(audio.Samples)) / float32(audio.SampleRate)
+
+	filename := fmt.Sprintf("seg-%d-%.2f-seconds-%s.wav", id, duration, name)
+	ok := audio.Save(filename)
+	if ok {
+		log.Printf("Saved to %s", filename)
+	}
+	log.Print("----------\n")
+}

go-api-examples/vad-speaker-identification/run.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+
+if [ ! -f ./3dspeaker_speech_campplus_sv_zh-cn_16k-common.onnx ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-recongition-models/3dspeaker_speech_campplus_sv_zh-cn_16k-common.onnx
+fi
+
+if [ ! -f ./sr-data/enroll/fangjun-sr-1.wav ]; then
+  git clone https://github.com/csukuangfj/sr-data
+fi
+
+go mod tidy
+go build
+./speaker-identification

go-api-examples/vad-spoken-language-identification/main.go

@@ -99,7 +99,7 @@ func main() {
 
 			duration := float32(len(speechSegment.Samples)) / float32(config.SampleRate)
 
-			audio := &sherpa.GeneratedAudio{}
+			audio := &sherpa.Wave{}
 			audio.Samples = speechSegment.Samples
 			audio.SampleRate = config.SampleRate
 
@@ -115,7 +115,7 @@ func main() {
 	chk(s.Stop())
 }
 
-func decode(slid *sherpa.SpokenLanguageIdentification, audio *sherpa.GeneratedAudio, id int) {
+func decode(slid *sherpa.SpokenLanguageIdentification, audio *sherpa.Wave, id int) {
 	stream := slid.CreateStream()
 	defer sherpa.DeleteOfflineStream(stream)
 

ios-swift/SherpaOnnx/SherpaOnnx.xcodeproj/xcuserdata/fangjun.xcuserdatad/xcschemes/xcschememanagement.plist

@@ -1,14 +0,0 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
-<plist version="1.0">
-<dict>
-	<key>SchemeUserState</key>
-	<dict>
-		<key>SherpaOnnx.xcscheme_^#shared#^_</key>
-		<dict>
-			<key>orderHint</key>
-			<integer>0</integer>
-		</dict>
-	</dict>
-</dict>
-</plist>

scripts/go/_internal/speaker-identification/.gitignore

@@ -0,0 +1 @@
+speaker-identification

scripts/go/_internal/speaker-identification/go.mod

@@ -0,0 +1,5 @@
+module speaker-identification
+
+go 1.12
+
+replace github.com/k2-fsa/sherpa-onnx-go/sherpa_onnx => ../

scripts/go/_internal/speaker-identification/main.go

@@ -0,0 +1 @@
+../../../../go-api-examples/speaker-identification/main.go

scripts/go/_internal/speaker-identification/run.sh

@@ -0,0 +1 @@
+../../../../go-api-examples/speaker-identification/run.sh

scripts/go/_internal/vad-asr-paraformer/go.mod

@@ -3,8 +3,3 @@ module vad-asr-paraformer
 go 1.12
 
 replace github.com/k2-fsa/sherpa-onnx-go/sherpa_onnx => ../
-
-require (
-	github.com/gordonklaus/portaudio v0.0.0-20230709114228-aafa478834f5
-	github.com/k2-fsa/sherpa-onnx-go/sherpa_onnx v0.0.0-00010101000000-000000000000
-)

scripts/go/_internal/vad-speaker-identification/go.mod

@@ -0,0 +1,5 @@
+module vad-speaker-identification
+
+go 1.12
+
+replace github.com/k2-fsa/sherpa-onnx-go/sherpa_onnx => ../

scripts/go/_internal/vad-speaker-identification/main.go

@@ -0,0 +1 @@
+../../../../go-api-examples/vad-speaker-identification/main.go

scripts/go/_internal/vad-speaker-identification/run.sh

@@ -0,0 +1 @@
+../../../../go-api-examples/vad-speaker-identification/run.sh

scripts/go/sherpa_onnx.go

@@ -746,11 +746,11 @@ func (vad *VoiceActivityDetector) AcceptWaveform(samples []float32) {
 }
 
 func (vad *VoiceActivityDetector) IsEmpty() bool {
-	return 1 == int(C.SherpaOnnxVoiceActivityDetectorEmpty(vad.impl))
+	return int(C.SherpaOnnxVoiceActivityDetectorEmpty(vad.impl)) == 1
 }
 
 func (vad *VoiceActivityDetector) IsSpeech() bool {
-	return 1 == int(C.SherpaOnnxVoiceActivityDetectorDetected(vad.impl))
+	return int(C.SherpaOnnxVoiceActivityDetectorDetected(vad.impl)) == 1
 }
 
 func (vad *VoiceActivityDetector) Pop() {
@@ -852,3 +852,204 @@ func (slid *SpokenLanguageIdentification) Compute(stream *OfflineStream) *Spoken
 
 	return ans
 }
+
+// ============================================================
+// For speaker embedding extraction
+// ============================================================
+
+type SpeakerEmbeddingExtractorConfig struct {
+	Model      string
+	NumThreads int
+	Debug      int
+	Provider   string
+}
+
+type SpeakerEmbeddingExtractor struct {
+	impl *C.struct_SherpaOnnxSpeakerEmbeddingExtractor
+}
+
+// The user has to invoke [DeleteSpeakerEmbeddingExtractor]() to free the returned value
+// to avoid memory leak
+func NewSpeakerEmbeddingExtractor(config *SpeakerEmbeddingExtractorConfig) *SpeakerEmbeddingExtractor {
+	c := C.struct_SherpaOnnxSpeakerEmbeddingExtractorConfig{}
+
+	c.model = C.CString(config.Model)
+	defer C.free(unsafe.Pointer(c.model))
+
+	c.num_threads = C.int(config.NumThreads)
+	c.debug = C.int(config.Debug)
+
+	c.provider = C.CString(config.Provider)
+	defer C.free(unsafe.Pointer(c.provider))
+
+	ex := &SpeakerEmbeddingExtractor{}
+	ex.impl = C.SherpaOnnxCreateSpeakerEmbeddingExtractor(&c)
+
+	return ex
+}
+
+func DeleteSpeakerEmbeddingExtractor(ex *SpeakerEmbeddingExtractor) {
+	C.SherpaOnnxDestroySpeakerEmbeddingExtractor(ex.impl)
+	ex.impl = nil
+}
+
+func (ex *SpeakerEmbeddingExtractor) Dim() int {
+	return int(C.SherpaOnnxSpeakerEmbeddingExtractorDim(ex.impl))
+}
+
+// The user is responsible to invoke [DeleteOnlineStream]() to free
+// the returned stream to avoid memory leak
+func (ex *SpeakerEmbeddingExtractor) CreateStream() *OnlineStream {
+	stream := &OnlineStream{}
+	stream.impl = C.SherpaOnnxSpeakerEmbeddingExtractorCreateStream(ex.impl)
+	return stream
+}
+
+func (ex *SpeakerEmbeddingExtractor) IsReady(stream *OnlineStream) bool {
+	return int(C.SherpaOnnxSpeakerEmbeddingExtractorIsReady(ex.impl, stream.impl)) == 1
+}
+
+func (ex *SpeakerEmbeddingExtractor) Compute(stream *OnlineStream) []float32 {
+	embedding := C.SherpaOnnxSpeakerEmbeddingExtractorComputeEmbedding(ex.impl, stream.impl)
+	defer C.SherpaOnnxSpeakerEmbeddingExtractorDestroyEmbedding(embedding)
+
+	n := ex.Dim()
+	ans := make([]float32, n)
+
+	// see https://stackoverflow.com/questions/48756732/what-does-1-30c-yourtype-do-exactly-in-cgo
+	// :n:n means 0:n:n, means low:high:capacity
+	c := (*[1 << 28]C.float)(unsafe.Pointer(embedding))[:n:n]
+
+	for i := 0; i < n; i++ {
+		ans[i] = float32(c[i])
+	}
+
+	return ans
+}
+
+type SpeakerEmbeddingManager struct {
+	impl *C.struct_SherpaOnnxSpeakerEmbeddingManager
+}
+
+// The user has to invoke [DeleteSpeakerEmbeddingManager]() to free the returned
+// value to avoid memory leak
+func NewSpeakerEmbeddingManager(dim int) *SpeakerEmbeddingManager {
+	m := &SpeakerEmbeddingManager{}
+	m.impl = C.SherpaOnnxCreateSpeakerEmbeddingManager(C.int(dim))
+	return m
+}
+
+func DeleteSpeakerEmbeddingManager(m *SpeakerEmbeddingManager) {
+	C.SherpaOnnxDestroySpeakerEmbeddingManager(m.impl)
+	m.impl = nil
+}
+
+func (m *SpeakerEmbeddingManager) Register(name string, embedding []float32) bool {
+	s := C.CString(name)
+	defer C.free(unsafe.Pointer(s))
+
+	return C.int(C.SherpaOnnxSpeakerEmbeddingManagerAdd(m.impl, s, (*C.float)(&embedding[0]))) == 1
+}
+
+func (m *SpeakerEmbeddingManager) RegisterV(name string, embeddings [][]float32) bool {
+	s := C.CString(name)
+	defer C.free(unsafe.Pointer(s))
+
+	if len(embeddings) == 0 {
+		return false
+	}
+
+	dim := len(embeddings[0])
+	v := make([]float32, 0, dim*len(embeddings))
+	for _, embedding := range embeddings {
+		v = append(v, embedding...)
+	}
+
+	return C.int(C.SherpaOnnxSpeakerEmbeddingManagerAddListFlattened(m.impl, s, (*C.float)(&v[0]), C.int(len(embeddings)))) == 1
+}
+
+func (m *SpeakerEmbeddingManager) Remove(name string) bool {
+	s := C.CString(name)
+	defer C.free(unsafe.Pointer(s))
+
+	return C.int(C.SherpaOnnxSpeakerEmbeddingManagerRemove(m.impl, s)) == 1
+}
+
+func (m *SpeakerEmbeddingManager) Search(embedding []float32, threshold float32) string {
+	var s string
+
+	name := C.SherpaOnnxSpeakerEmbeddingManagerSearch(m.impl, (*C.float)(&embedding[0]), C.float(threshold))
+	defer C.SherpaOnnxSpeakerEmbeddingManagerFreeSearch(name)
+
+	if name != nil {
+		s = C.GoString(name)
+	}
+
+	return s
+}
+
+func (m *SpeakerEmbeddingManager) Verify(name string, embedding []float32, threshold float32) bool {
+	s := C.CString(name)
+	defer C.free(unsafe.Pointer(s))
+
+	return C.int(C.SherpaOnnxSpeakerEmbeddingManagerVerify(m.impl, s, (*C.float)(&embedding[0]), C.float(threshold))) == 1
+}
+
+func (m *SpeakerEmbeddingManager) Contains(name string) bool {
+	s := C.CString(name)
+	defer C.free(unsafe.Pointer(s))
+
+	return C.int(C.SherpaOnnxSpeakerEmbeddingManagerContains(m.impl, s)) == 1
+}
+
+func (m *SpeakerEmbeddingManager) NumSpeakers() int {
+	return int(C.SherpaOnnxSpeakerEmbeddingManagerNumSpeakers(m.impl))
+}
+
+func (m *SpeakerEmbeddingManager) AllSpeakers() []string {
+	all_speakers := C.SherpaOnnxSpeakerEmbeddingManagerGetAllSpeakers(m.impl)
+	defer C.SherpaOnnxSpeakerEmbeddingManagerFreeAllSpeakers(all_speakers)
+
+	n := m.NumSpeakers()
+	if n == 0 {
+		return nil
+	}
+
+	// https://stackoverflow.com/questions/62012070/convert-array-of-strings-from-cgo-in-go
+	p := (*[1 << 28]*C.char)(unsafe.Pointer(all_speakers))[:n:n]
+
+	ans := make([]string, n)
+
+	for i := 0; i < n; i++ {
+		ans[i] = C.GoString(p[i])
+	}
+
+	return ans
+}
+
+// Wave
+
+// single channel wave
+type Wave = GeneratedAudio
+
+func ReadWave(filename string) *Wave {
+	s := C.CString(filename)
+	defer C.free(unsafe.Pointer(s))
+
+	w := C.SherpaOnnxReadWave(s)
+	defer C.SherpaOnnxFreeWave(w)
+
+	n := int(w.num_samples)
+
+	ans := &Wave{}
+	ans.SampleRate = int(w.sample_rate)
+	samples := (*[1 << 28]C.float)(unsafe.Pointer(w.samples))[:n:n]
+
+	ans.Samples = make([]float32, n)
+
+	for i := 0; i < n; i++ {
+		ans.Samples[i] = float32(samples[i])
+	}
+
+	return ans
+}