sherpa_onnx

package module
v0.0.0-...-4d8d542 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: May 13, 2025 License: Apache-2.0 Imports: 2 Imported by: 0

README

Introduction

This repo contains the Go package of sherpa-onnx for macOS, supporting x86_64 (Intel chips) as well as aarch64 (Apple Silicon, e.g., M1).

Documentation

Overview

Speech recognition with Next-gen Kaldi.

sherpa-onnx is an open-source speech recognition framework for Next-gen Kaldi. It depends only on onnxruntime, supporting both streaming and non-streaming speech recognition.

It does not need to access the network during recognition and everything runs locally.

It supports a variety of platforms, such as Linux (x86_64, aarch64, arm), Windows (x86_64, x86), macOS (x86_64, arm64), etc.

Usage examples:

  1. Real-time speech recognition from a microphone

    Please see https://github.com/k2-fsa/sherpa-onnx/tree/master/go-api-examples/real-time-speech-recognition-from-microphone

  2. Decode files using a non-streaming model

    Please see https://github.com/k2-fsa/sherpa-onnx/tree/master/go-api-examples/non-streaming-decode-files

  3. Decode files using a streaming model

    Please see https://github.com/k2-fsa/sherpa-onnx/tree/master/go-api-examples/streaming-decode-files

  4. Convert text to speech using a non-streaming model

    Please see https://github.com/k2-fsa/sherpa-onnx/tree/master/go-api-examples/non-streaming-tts

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

func DeleteAudioTagging 

func DeleteAudioTagging(tagging *AudioTagging)

func DeleteCircularBuffer 

func DeleteCircularBuffer(buffer *CircularBuffer)

func DeleteKeywordSpotter 

func DeleteKeywordSpotter(spotter *KeywordSpotter)

Free the internal pointer inside the recognizer to avoid memory leak.

func DeleteOfflinePunc 

func DeleteOfflinePunc(punc *OfflinePunctuation)

func DeleteOfflineRecognizer 

func DeleteOfflineRecognizer(recognizer *OfflineRecognizer)

Frees the internal pointer of the recognition to avoid memory leak.

func DeleteOfflineSpeakerDiarization 

func DeleteOfflineSpeakerDiarization(sd *OfflineSpeakerDiarization)

func DeleteOfflineSpeechDenoiser 

func DeleteOfflineSpeechDenoiser(sd *OfflineSpeechDenoiser)

Free the internal pointer inside the OfflineSpeechDenoiser to avoid memory leak.

func DeleteOfflineStream 

func DeleteOfflineStream(stream *OfflineStream)

Frees the internal pointer of the stream to avoid memory leak.

func DeleteOfflineTts 

func DeleteOfflineTts(tts *OfflineTts)

Free the internal pointer inside the tts to avoid memory leak.

func DeleteOnlineRecognizer 

func DeleteOnlineRecognizer(recognizer *OnlineRecognizer)

Free the internal pointer inside the recognizer to avoid memory leak.

func DeleteOnlineStream 

func DeleteOnlineStream(stream *OnlineStream)

Delete the internal pointer inside the stream to avoid memory leak.

func DeleteSpeakerEmbeddingExtractor 

func DeleteSpeakerEmbeddingExtractor(ex *SpeakerEmbeddingExtractor)

func DeleteSpeakerEmbeddingManager 

func DeleteSpeakerEmbeddingManager(m *SpeakerEmbeddingManager)

func DeleteSpokenLanguageIdentification 

func DeleteSpokenLanguageIdentification(slid *SpokenLanguageIdentification)

func DeleteVoiceActivityDetector 

func DeleteVoiceActivityDetector(vad *VoiceActivityDetector)

Types

type AudioEvent 

type AudioEvent struct {
	Name  string
	Index int
	Prob  float32
}

type AudioTagging 

type AudioTagging struct {
	// contains filtered or unexported fields
}

func NewAudioTagging 

func NewAudioTagging(config *AudioTaggingConfig) *AudioTagging

The user is responsible to invoke DeleteAudioTagging() to free the returned tagger to avoid memory leak

func (*AudioTagging) Compute 

func (tagging *AudioTagging) Compute(s *OfflineStream, topK int32) []AudioEvent

type AudioTaggingConfig 

type AudioTaggingConfig struct {
	Model  AudioTaggingModelConfig
	Labels string
	TopK   int32
}

type AudioTaggingModelConfig 

type AudioTaggingModelConfig struct {
	Zipformer  OfflineZipformerAudioTaggingModelConfig
	Ced        string
	NumThreads int32
	Debug      int32
	Provider   string
}

type CircularBuffer 

type CircularBuffer struct {
	// contains filtered or unexported fields
}

func NewCircularBuffer 

func NewCircularBuffer(capacity int) *CircularBuffer

func (*CircularBuffer) Get 

func (buffer *CircularBuffer) Get(start int, n int) []float32

func (*CircularBuffer) Head 

func (buffer *CircularBuffer) Head() int

func (*CircularBuffer) Pop 

func (buffer *CircularBuffer) Pop(n int)

func (*CircularBuffer) Push 

func (buffer *CircularBuffer) Push(samples []float32)

func (*CircularBuffer) Reset 

func (buffer *CircularBuffer) Reset()

func (*CircularBuffer) Size 

func (buffer *CircularBuffer) Size() int

type DenoisedAudio 

type DenoisedAudio struct {
	// Normalized samples in the range [-1, 1]
	Samples []float32

	SampleRate int
}

func (*DenoisedAudio) Save 

func (audio *DenoisedAudio) Save(filename string) bool

type FastClusteringConfig 

type FastClusteringConfig struct {
	NumClusters int
	Threshold   float32
}

type FeatureConfig 

type FeatureConfig struct {
	// Sample rate expected by the model. It is 16000 for all
	// pre-trained models provided by us
	SampleRate int
	// Feature dimension expected by the model. It is 80 for all
	// pre-trained models provided by us
	FeatureDim int
}

Configuration for the feature extractor

type GeneratedAudio 

type GeneratedAudio struct {
	// Normalized samples in the range [-1, 1]
	Samples []float32

	SampleRate int
}

func (*GeneratedAudio) Save 

func (audio *GeneratedAudio) Save(filename string) bool

type HomophoneReplacerConfig 

type HomophoneReplacerConfig struct {
	DictDir  string
	Lexicon  string
	RuleFsts string
}

type KeywordSpotter 

type KeywordSpotter struct {
	// contains filtered or unexported fields
}

func NewKeywordSpotter 

func NewKeywordSpotter(config *KeywordSpotterConfig) *KeywordSpotter

The user is responsible to invoke DeleteKeywordSpotter() to free the returned spotter to avoid memory leak

func (*KeywordSpotter) Decode 

func (spotter *KeywordSpotter) Decode(s *OnlineStream)

Decode the stream. Before calling this function, you have to ensure that spotter.IsReady(s) returns true. Otherwise, you will be SAD.

You usually use it like below: 

for spotter.IsReady(s) {
  spotter.Decode(s)
}

func (*KeywordSpotter) GetResult 

func (spotter *KeywordSpotter) GetResult(s *OnlineStream) *KeywordSpotterResult

Get the current result of stream since the last invoke of Reset()

func (*KeywordSpotter) IsReady 

func (spotter *KeywordSpotter) IsReady(s *OnlineStream) bool

Check whether the stream has enough feature frames for decoding. Return true if this stream is ready for decoding. Return false otherwise.

You will usually use it like below: 

for spotter.IsReady(s) {
   spotter.Decode(s)
}

func (*KeywordSpotter) Reset 

func (spotter *KeywordSpotter) Reset(s *OnlineStream)

You MUST call it right after detecting a keyword

type KeywordSpotterConfig 

type KeywordSpotterConfig struct {
	FeatConfig        FeatureConfig
	ModelConfig       OnlineModelConfig
	MaxActivePaths    int
	KeywordsFile      string
	KeywordsScore     float32
	KeywordsThreshold float32
	KeywordsBuf       string
	KeywordsBufSize   int
}

Configuration for the online/streaming recognizer.

type KeywordSpotterResult 

type KeywordSpotterResult struct {
	Keyword string
}

type OfflineDolphinModelConfig 

type OfflineDolphinModelConfig struct {
	Model string // Path to the model, e.g., model.onnx or model.int8.onnx
}

type OfflineFireRedAsrModelConfig 

type OfflineFireRedAsrModelConfig struct {
	Encoder string
	Decoder string
}

type OfflineLMConfig 

type OfflineLMConfig struct {
	Model string  // Path to the model
	Scale float32 // scale for LM score
}

Configuration for offline LM.

type OfflineModelConfig 

type OfflineModelConfig struct {
	Transducer OfflineTransducerModelConfig
	Paraformer OfflineParaformerModelConfig
	NemoCTC    OfflineNemoEncDecCtcModelConfig
	Whisper    OfflineWhisperModelConfig
	Tdnn       OfflineTdnnModelConfig
	SenseVoice OfflineSenseVoiceModelConfig
	Moonshine  OfflineMoonshineModelConfig
	FireRedAsr OfflineFireRedAsrModelConfig
	Dolphin    OfflineDolphinModelConfig
	Tokens     string // Path to tokens.txt

	// Number of threads to use for neural network computation
	NumThreads int

	// 1 to print model meta information while loading
	Debug int

	// Optional. Valid values: cpu, cuda, coreml
	Provider string

	// Optional. Specify it for faster model initialization.
	ModelType string

	ModelingUnit  string // Optional. cjkchar, bpe, cjkchar+bpe
	BpeVocab      string // Optional.
	TeleSpeechCtc string // Optional.
}

type OfflineMoonshineModelConfig 

type OfflineMoonshineModelConfig struct {
	Preprocessor    string
	Encoder         string
	UncachedDecoder string
	CachedDecoder   string
}

type OfflineNemoEncDecCtcModelConfig 

type OfflineNemoEncDecCtcModelConfig struct {
	Model string // Path to the model, e.g., model.onnx or model.int8.onnx
}

Configuration for offline/non-streaming NeMo CTC models.

Please refer to https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-ctc/index.html to download pre-trained models

type OfflineParaformerModelConfig 

type OfflineParaformerModelConfig struct {
	Model string // Path to the model, e.g., model.onnx or model.int8.onnx
}

Configuration for offline/non-streaming paraformer.

please refer to https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-paraformer/index.html to download pre-trained models

type OfflinePunctuation 

type OfflinePunctuation struct {
	// contains filtered or unexported fields
}

func NewOfflinePunctuation 

func NewOfflinePunctuation(config *OfflinePunctuationConfig) *OfflinePunctuation

func (*OfflinePunctuation) AddPunct 

func (punc *OfflinePunctuation) AddPunct(text string) string

type OfflinePunctuationConfig 

type OfflinePunctuationConfig struct {
	Model OfflinePunctuationModelConfig
}

type OfflinePunctuationModelConfig 

type OfflinePunctuationModelConfig struct {
	CtTransformer string
	NumThreads    C.int
	Debug         C.int // true to print debug information of the model
	Provider      string
}

============================================================ For punctuation ============================================================

type OfflineRecognizer 

type OfflineRecognizer struct {
	// contains filtered or unexported fields
}

It wraps a pointer from C

func NewOfflineRecognizer 

func NewOfflineRecognizer(config *OfflineRecognizerConfig) *OfflineRecognizer

The user is responsible to invoke DeleteOfflineRecognizer() to free the returned recognizer to avoid memory leak

func (*OfflineRecognizer) Decode 

func (recognizer *OfflineRecognizer) Decode(s *OfflineStream)

Decode the offline stream.

func (*OfflineRecognizer) DecodeStreams 

func (recognizer *OfflineRecognizer) DecodeStreams(s []*OfflineStream)

Decode multiple streams in parallel, i.e., in batch.

func (*OfflineRecognizer) SetConfig 

func (r *OfflineRecognizer) SetConfig(config *OfflineRecognizerConfig)

Set new config to replace

type OfflineRecognizerConfig 

type OfflineRecognizerConfig struct {
	FeatConfig  FeatureConfig
	ModelConfig OfflineModelConfig
	LmConfig    OfflineLMConfig

	// Valid decoding method: greedy_search, modified_beam_search
	DecodingMethod string

	// Used only when DecodingMethod is modified_beam_search.
	MaxActivePaths int
	HotwordsFile   string
	HotwordsScore  float32
	BlankPenalty   float32
	RuleFsts       string
	RuleFars       string
	Hr             HomophoneReplacerConfig
}

Configuration for the offline/non-streaming recognizer.

type OfflineRecognizerResult 

type OfflineRecognizerResult struct {
	Text       string
	Tokens     []string
	Timestamps []float32
	Lang       string
	Emotion    string
	Event      string
}

It contains recognition result of an offline stream.

type OfflineSenseVoiceModelConfig 

type OfflineSenseVoiceModelConfig struct {
	Model                       string
	Language                    string
	UseInverseTextNormalization int
}

type OfflineSpeakerDiarization 

type OfflineSpeakerDiarization struct {
	// contains filtered or unexported fields
}

func NewOfflineSpeakerDiarization 

func NewOfflineSpeakerDiarization(config *OfflineSpeakerDiarizationConfig) *OfflineSpeakerDiarization

func (*OfflineSpeakerDiarization) Process 

func (sd *OfflineSpeakerDiarization) Process(samples []float32) []OfflineSpeakerDiarizationSegment

func (*OfflineSpeakerDiarization) SampleRate 

func (sd *OfflineSpeakerDiarization) SampleRate() int

func (*OfflineSpeakerDiarization) SetConfig 

func (sd *OfflineSpeakerDiarization) SetConfig(config *OfflineSpeakerDiarizationConfig)

only config.Clustering is used. All other fields are ignored

type OfflineSpeakerDiarizationConfig 

type OfflineSpeakerDiarizationConfig struct {
	Segmentation   OfflineSpeakerSegmentationModelConfig
	Embedding      SpeakerEmbeddingExtractorConfig
	Clustering     FastClusteringConfig
	MinDurationOn  float32
	MinDurationOff float32
}

type OfflineSpeakerDiarizationSegment 

type OfflineSpeakerDiarizationSegment struct {
	Start   float32
	End     float32
	Speaker int
}

type OfflineSpeakerSegmentationModelConfig 

type OfflineSpeakerSegmentationModelConfig struct {
	Pyannote   OfflineSpeakerSegmentationPyannoteModelConfig
	NumThreads int
	Debug      int
	Provider   string
}

type OfflineSpeakerSegmentationPyannoteModelConfig 

type OfflineSpeakerSegmentationPyannoteModelConfig struct {
	Model string
}

============================================================ For offline speaker diarization ============================================================

type OfflineSpeechDenoiser 

type OfflineSpeechDenoiser struct {
	// contains filtered or unexported fields
}

func NewOfflineSpeechDenoiser 

func NewOfflineSpeechDenoiser(config *OfflineSpeechDenoiserConfig) *OfflineSpeechDenoiser

The user is responsible to invoke DeleteOfflineSpeechDenoiser() to free the returned tts to avoid memory leak

func (*OfflineSpeechDenoiser) Run 

func (sd *OfflineSpeechDenoiser) Run(samples []float32, sampleRate int) *DenoisedAudio

func (*OfflineSpeechDenoiser) SampleRate 

func (sd *OfflineSpeechDenoiser) SampleRate() int

type OfflineSpeechDenoiserConfig 

type OfflineSpeechDenoiserConfig struct {
	Model OfflineSpeechDenoiserModelConfig
}

type OfflineSpeechDenoiserGtcrnModelConfig 

type OfflineSpeechDenoiserGtcrnModelConfig struct {
	Model string
}

type OfflineSpeechDenoiserModelConfig 

type OfflineSpeechDenoiserModelConfig struct {
	Gtcrn      OfflineSpeechDenoiserGtcrnModelConfig
	NumThreads int32
	Debug      int32
	Provider   string
}

type OfflineStream 

type OfflineStream struct {
	// contains filtered or unexported fields
}

It wraps a pointer from C

func NewAudioTaggingStream 

func NewAudioTaggingStream(tagging *AudioTagging) *OfflineStream

The user is responsible to invoke DeleteOfflineStream() to free the returned stream to avoid memory leak

func NewOfflineStream 

func NewOfflineStream(recognizer *OfflineRecognizer) *OfflineStream

The user is responsible to invoke DeleteOfflineStream() to free the returned stream to avoid memory leak

func (*OfflineStream) AcceptWaveform 

func (s *OfflineStream) AcceptWaveform(sampleRate int, samples []float32)

Input audio samples for the offline stream. Please only call it once. That is, input all samples at once.

sampleRate is the sample rate of the input audio samples. If it is different from the value expected by the feature extractor, we will do resampling inside.

samples contains the actual audio samples. Each sample is in the range [-1, 1].

func (*OfflineStream) GetResult 

func (s *OfflineStream) GetResult() *OfflineRecognizerResult

Get the recognition result of the offline stream.

type OfflineTdnnModelConfig 

type OfflineTdnnModelConfig struct {
	Model string
}

type OfflineTransducerModelConfig 

type OfflineTransducerModelConfig struct {
	Encoder string // Path to the encoder model, i.e., encoder.onnx or encoder.int8.onnx
	Decoder string // Path to the decoder model
	Joiner  string // Path to the joiner model
}

Configuration for offline/non-streaming transducer.

Please refer to https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-transducer/index.html to download pre-trained models

type OfflineTts 

type OfflineTts struct {
	// contains filtered or unexported fields
}

The offline tts class. It wraps a pointer from C.

func NewOfflineTts 

func NewOfflineTts(config *OfflineTtsConfig) *OfflineTts

The user is responsible to invoke DeleteOfflineTts() to free the returned tts to avoid memory leak

func (*OfflineTts) Generate 

func (tts *OfflineTts) Generate(text string, sid int, speed float32) *GeneratedAudio

type OfflineTtsConfig 

type OfflineTtsConfig struct {
	Model           OfflineTtsModelConfig
	RuleFsts        string
	RuleFars        string
	MaxNumSentences int
	SilenceScale    float32
}

type OfflineTtsKokoroModelConfig 

type OfflineTtsKokoroModelConfig struct {
	Model       string  // Path to the model for kokoro
	Voices      string  // Path to the voices.bin for kokoro
	Tokens      string  // Path to tokens.txt
	DataDir     string  // Path to espeak-ng-data directory
	DictDir     string  // Path to dict directory
	Lexicon     string  // Path to lexicon files
	LengthScale float32 // Please use 1.0 in general. Smaller -> Faster speech speed. Larger -> Slower speech speed
}

type OfflineTtsMatchaModelConfig 

type OfflineTtsMatchaModelConfig struct {
	AcousticModel string  // Path to the acoustic model for MatchaTTS
	Vocoder       string  // Path to the vocoder model for MatchaTTS
	Lexicon       string  // Path to lexicon.txt
	Tokens        string  // Path to tokens.txt
	DataDir       string  // Path to espeak-ng-data directory
	NoiseScale    float32 // noise scale for vits models. Please use 0.667 in general
	LengthScale   float32 // Please use 1.0 in general. Smaller -> Faster speech speed. Larger -> Slower speech speed
	DictDir       string  // Path to dict directory for jieba (used only in Chinese tts)
}

type OfflineTtsModelConfig 

type OfflineTtsModelConfig struct {
	Vits   OfflineTtsVitsModelConfig
	Matcha OfflineTtsMatchaModelConfig
	Kokoro OfflineTtsKokoroModelConfig

	// Number of threads to use for neural network computation
	NumThreads int

	// 1 to print model meta information while loading
	Debug int

	// Optional. Valid values: cpu, cuda, coreml
	Provider string
}

type OfflineTtsVitsModelConfig 

type OfflineTtsVitsModelConfig struct {
	Model       string  // Path to the VITS onnx model
	Lexicon     string  // Path to lexicon.txt
	Tokens      string  // Path to tokens.txt
	DataDir     string  // Path to espeak-ng-data directory
	NoiseScale  float32 // noise scale for vits models. Please use 0.667 in general
	NoiseScaleW float32 // noise scale for vits models. Please use 0.8 in general
	LengthScale float32 // Please use 1.0 in general. Smaller -> Faster speech speed. Larger -> Slower speech speed
	DictDir     string  // Path to dict directory for jieba (used only in Chinese tts)
}

Configuration for offline/non-streaming text-to-speech (TTS).

Please refer to https://k2-fsa.github.io/sherpa/onnx/tts/pretrained_models/index.html to download pre-trained models

type OfflineWhisperModelConfig 

type OfflineWhisperModelConfig struct {
	Encoder      string
	Decoder      string
	Language     string
	Task         string
	TailPaddings int
}

type OfflineZipformerAudioTaggingModelConfig 

type OfflineZipformerAudioTaggingModelConfig struct {
	Model string
}

Configuration for the audio tagging.

type OnlineCtcFstDecoderConfig 

type OnlineCtcFstDecoderConfig struct {
	Graph     string
	MaxActive int
}

type OnlineModelConfig 

type OnlineModelConfig struct {
	Transducer    OnlineTransducerModelConfig
	Paraformer    OnlineParaformerModelConfig
	Zipformer2Ctc OnlineZipformer2CtcModelConfig
	Tokens        string // Path to tokens.txt
	NumThreads    int    // Number of threads to use for neural network computation
	Provider      string // Optional. Valid values are: cpu, cuda, coreml
	Debug         int    // 1 to show model meta information while loading it.
	ModelType     string // Optional. You can specify it for faster model initialization
	ModelingUnit  string // Optional. cjkchar, bpe, cjkchar+bpe
	BpeVocab      string // Optional.
	TokensBuf     string // Optional.
	TokensBufSize int    // Optional.
}

Configuration for online/streaming models

Please refer to https://k2-fsa.github.io/sherpa/onnx/pretrained_models/online-transducer/index.html https://k2-fsa.github.io/sherpa/onnx/pretrained_models/online-paraformer/index.html to download pre-trained models

type OnlineParaformerModelConfig 

type OnlineParaformerModelConfig struct {
	Encoder string // Path to the encoder model, e.g., encoder.onnx or encoder.int8.onnx
	Decoder string // Path to the decoder model.
}

Configuration for online/streaming paraformer models

Please refer to https://k2-fsa.github.io/sherpa/onnx/pretrained_models/online-paraformer/index.html to download pre-trained models

type OnlineRecognizer 

type OnlineRecognizer struct {
	// contains filtered or unexported fields
}

The online recognizer class. It wraps a pointer from C.

func NewOnlineRecognizer 

func NewOnlineRecognizer(config *OnlineRecognizerConfig) *OnlineRecognizer

The user is responsible to invoke DeleteOnlineRecognizer() to free the returned recognizer to avoid memory leak

func (*OnlineRecognizer) Decode 

func (recognizer *OnlineRecognizer) Decode(s *OnlineStream)

Decode the stream. Before calling this function, you have to ensure that recognizer.IsReady(s) returns true. Otherwise, you will be SAD.

You usually use it like below: 

for recognizer.IsReady(s) {
  recognizer.Decode(s)
}

func (*OnlineRecognizer) DecodeStreams 

func (recognizer *OnlineRecognizer) DecodeStreams(s []*OnlineStream)

Decode multiple streams in parallel, i.e., in batch. You have to ensure that each stream is ready for decoding. Otherwise, you will be SAD.

func (*OnlineRecognizer) GetResult 

func (recognizer *OnlineRecognizer) GetResult(s *OnlineStream) *OnlineRecognizerResult

Get the current result of stream since the last invoke of Reset()

func (*OnlineRecognizer) IsEndpoint 

func (recognizer *OnlineRecognizer) IsEndpoint(s *OnlineStream) bool

Return true if an endpoint is detected.

You usually use it like below: 

if recognizer.IsEndpoint(s) {
   // do your own stuff after detecting an endpoint

   recognizer.Reset(s)
}

func (*OnlineRecognizer) IsReady 

func (recognizer *OnlineRecognizer) IsReady(s *OnlineStream) bool

Check whether the stream has enough feature frames for decoding. Return true if this stream is ready for decoding. Return false otherwise.

You will usually use it like below: 

for recognizer.IsReady(s) {
   recognizer.Decode(s)
}

func (*OnlineRecognizer) Reset 

func (recognizer *OnlineRecognizer) Reset(s *OnlineStream)

After calling this function, the internal neural network model states are reset and IsEndpoint(s) would return false. GetResult(s) would also return an empty string.

type OnlineRecognizerConfig 

type OnlineRecognizerConfig struct {
	FeatConfig  FeatureConfig
	ModelConfig OnlineModelConfig

	// Valid decoding methods: greedy_search, modified_beam_search
	DecodingMethod string

	// Used only when DecodingMethod is modified_beam_search. It specifies
	// the maximum number of paths to keep during the search
	MaxActivePaths int

	EnableEndpoint int // 1 to enable endpoint detection.

	// Please see
	// https://k2-fsa.github.io/sherpa/ncnn/endpoint.html
	// for the meaning of Rule1MinTrailingSilence, Rule2MinTrailingSilence
	// and Rule3MinUtteranceLength.
	Rule1MinTrailingSilence float32
	Rule2MinTrailingSilence float32
	Rule3MinUtteranceLength float32
	HotwordsFile            string
	HotwordsScore           float32
	BlankPenalty            float32
	CtcFstDecoderConfig     OnlineCtcFstDecoderConfig
	RuleFsts                string
	RuleFars                string
	HotwordsBuf             string
	HotwordsBufSize         int
	Hr                      HomophoneReplacerConfig
}

Configuration for the online/streaming recognizer.

type OnlineRecognizerResult 

type OnlineRecognizerResult struct {
	Text string
}

It contains the recognition result for a online stream.

type OnlineStream 

type OnlineStream struct {
	// contains filtered or unexported fields
}

The online stream class. It wraps a pointer from C.

func NewKeywordStream 

func NewKeywordStream(spotter *KeywordSpotter) *OnlineStream

The user is responsible to invoke DeleteOnlineStream() to free the returned stream to avoid memory leak

func NewKeywordStreamWithKeywords 

func NewKeywordStreamWithKeywords(spotter *KeywordSpotter, keywords string) *OnlineStream

The user is responsible to invoke DeleteOnlineStream() to free the returned stream to avoid memory leak

func NewOnlineStream 

func NewOnlineStream(recognizer *OnlineRecognizer) *OnlineStream

The user is responsible to invoke DeleteOnlineStream() to free the returned stream to avoid memory leak

func (*OnlineStream) AcceptWaveform 

func (s *OnlineStream) AcceptWaveform(sampleRate int, samples []float32)

Input audio samples for the stream.

sampleRate is the actual sample rate of the input audio samples. If it is different from the sample rate expected by the feature extractor, we will do resampling inside.

samples contains audio samples. Each sample is in the range [-1, 1]

func (*OnlineStream) InputFinished 

func (s *OnlineStream) InputFinished()

Signal that there will be no incoming audio samples. After calling this function, you cannot call OnlineStream.AcceptWaveform any longer.

The main purpose of this function is to flush the remaining audio samples buffered inside for feature extraction.

type OnlineTransducerModelConfig 

type OnlineTransducerModelConfig struct {
	Encoder string // Path to the encoder model, e.g., encoder.onnx or encoder.int8.onnx
	Decoder string // Path to the decoder model.
	Joiner  string // Path to the joiner model.
}

Configuration for online/streaming transducer models

Please refer to https://k2-fsa.github.io/sherpa/onnx/pretrained_models/online-transducer/index.html to download pre-trained models

type OnlineZipformer2CtcModelConfig 

type OnlineZipformer2CtcModelConfig struct {
	Model string // Path to the onnx model
}

Please refer to https://k2-fsa.github.io/sherpa/onnx/pretrained_models/online-ctc/index.html to download pre-trained models

type SileroVadModelConfig 

type SileroVadModelConfig struct {
	Model              string
	Threshold          float32
	MinSilenceDuration float32
	MinSpeechDuration  float32
	WindowSize         int
	MaxSpeechDuration  float32
}

============================================================ For VAD ============================================================

type SpeakerEmbeddingExtractor 

type SpeakerEmbeddingExtractor struct {
	// contains filtered or unexported fields
}

func NewSpeakerEmbeddingExtractor 

func NewSpeakerEmbeddingExtractor(config *SpeakerEmbeddingExtractorConfig) *SpeakerEmbeddingExtractor

The user has to invoke DeleteSpeakerEmbeddingExtractor() to free the returned value to avoid memory leak

func (*SpeakerEmbeddingExtractor) Compute 

func (ex *SpeakerEmbeddingExtractor) Compute(stream *OnlineStream) []float32

func (*SpeakerEmbeddingExtractor) CreateStream 

func (ex *SpeakerEmbeddingExtractor) CreateStream() *OnlineStream

The user is responsible to invoke DeleteOnlineStream() to free the returned stream to avoid memory leak

func (*SpeakerEmbeddingExtractor) Dim 

func (ex *SpeakerEmbeddingExtractor) Dim() int

func (*SpeakerEmbeddingExtractor) IsReady 

func (ex *SpeakerEmbeddingExtractor) IsReady(stream *OnlineStream) bool

type SpeakerEmbeddingExtractorConfig 

type SpeakerEmbeddingExtractorConfig struct {
	Model      string
	NumThreads int
	Debug      int
	Provider   string
}

type SpeakerEmbeddingManager 

type SpeakerEmbeddingManager struct {
	// contains filtered or unexported fields
}

func NewSpeakerEmbeddingManager 

func NewSpeakerEmbeddingManager(dim int) *SpeakerEmbeddingManager

The user has to invoke DeleteSpeakerEmbeddingManager() to free the returned value to avoid memory leak

func (*SpeakerEmbeddingManager) AllSpeakers 

func (m *SpeakerEmbeddingManager) AllSpeakers() []string

func (*SpeakerEmbeddingManager) Contains 

func (m *SpeakerEmbeddingManager) Contains(name string) bool

func (*SpeakerEmbeddingManager) NumSpeakers 

func (m *SpeakerEmbeddingManager) NumSpeakers() int

func (*SpeakerEmbeddingManager) Register 

func (m *SpeakerEmbeddingManager) Register(name string, embedding []float32) bool

func (*SpeakerEmbeddingManager) RegisterV 

func (m *SpeakerEmbeddingManager) RegisterV(name string, embeddings [][]float32) bool

func (*SpeakerEmbeddingManager) Remove 

func (m *SpeakerEmbeddingManager) Remove(name string) bool

func (*SpeakerEmbeddingManager) Search 

func (m *SpeakerEmbeddingManager) Search(embedding []float32, threshold float32) string

func (*SpeakerEmbeddingManager) Verify 

func (m *SpeakerEmbeddingManager) Verify(name string, embedding []float32, threshold float32) bool

type SpeechSegment 

type SpeechSegment struct {
	Start   int
	Samples []float32
}

type SpokenLanguageIdentification 

type SpokenLanguageIdentification struct {
	// contains filtered or unexported fields
}

func NewSpokenLanguageIdentification 

func NewSpokenLanguageIdentification(config *SpokenLanguageIdentificationConfig) *SpokenLanguageIdentification

func (*SpokenLanguageIdentification) Compute 

func (slid *SpokenLanguageIdentification) Compute(stream *OfflineStream) *SpokenLanguageIdentificationResult

func (*SpokenLanguageIdentification) CreateStream 

func (slid *SpokenLanguageIdentification) CreateStream() *OfflineStream

The user has to invoke DeleteOfflineStream() to free the returned value to avoid memory leak

type SpokenLanguageIdentificationConfig 

type SpokenLanguageIdentificationConfig struct {
	Whisper    SpokenLanguageIdentificationWhisperConfig
	NumThreads int
	Debug      int
	Provider   string
}

type SpokenLanguageIdentificationResult 

type SpokenLanguageIdentificationResult struct {
	Lang string
}

type SpokenLanguageIdentificationWhisperConfig 

type SpokenLanguageIdentificationWhisperConfig struct {
	Encoder      string
	Decoder      string
	TailPaddings int
}

type VadModelConfig 

type VadModelConfig struct {
	SileroVad  SileroVadModelConfig
	SampleRate int
	NumThreads int
	Provider   string
	Debug      int
}

type VoiceActivityDetector 

type VoiceActivityDetector struct {
	// contains filtered or unexported fields
}

func NewVoiceActivityDetector 

func NewVoiceActivityDetector(config *VadModelConfig, bufferSizeInSeconds float32) *VoiceActivityDetector

func (*VoiceActivityDetector) AcceptWaveform 

func (vad *VoiceActivityDetector) AcceptWaveform(samples []float32)

func (*VoiceActivityDetector) Clear 

func (vad *VoiceActivityDetector) Clear()

func (*VoiceActivityDetector) Flush 

func (vad *VoiceActivityDetector) Flush()

func (*VoiceActivityDetector) Front 

func (vad *VoiceActivityDetector) Front() *SpeechSegment

func (*VoiceActivityDetector) IsEmpty 

func (vad *VoiceActivityDetector) IsEmpty() bool

func (*VoiceActivityDetector) IsSpeech 

func (vad *VoiceActivityDetector) IsSpeech() bool

func (*VoiceActivityDetector) Pop 

func (vad *VoiceActivityDetector) Pop()

func (*VoiceActivityDetector) Reset 

func (vad *VoiceActivityDetector) Reset()

type Wave 

type Wave = GeneratedAudio

single channel wave

func ReadWave 

func ReadWave(filename string) *Wave

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.