vosk

README

See example subfolder for instructions how to use the module

Documentation

Overview

Go bindings for Vosk speech recognition toolkit. Vosk is an offline open source speech to text API for Android, iOS, Raspberry Pi and servers. It enables speech recognition models for 18 languages and dialects - English, Indian English, German, French, Spanish, Portuguese, Chinese, Russian, Turkish, Vietnamese, Italian, Dutch, Catalan, Arabic, Greek, Farsi, Filipino, Ukrainian.

Index

• func GPUInit()
• func GPUThreadInit()
• func SetLogLevel(logLevel int)
• type VoskModel
  • func NewModel(modelPath string) (*VoskModel, error)
  • func (m *VoskModel) FindWord(word string) int
  • func (m *VoskModel) Free()
• type VoskRecognizer
  • func NewRecognizer(model *VoskModel, sampleRate float64) (*VoskRecognizer, error)
  • func NewRecognizerGrm(model *VoskModel, sampleRate float64, grammar string) (*VoskRecognizer, error)
  • func NewRecognizerSpk(model *VoskModel, sampleRate float64, spkModel *VoskSpkModel) (*VoskRecognizer, error)
  • func (r *VoskRecognizer) AcceptWaveform(buffer []byte) int
  • func (r *VoskRecognizer) FinalResult() string
  • func (r *VoskRecognizer) Free()
  • func (r *VoskRecognizer) PartialResult() string
  • func (r *VoskRecognizer) Reset()
  • func (r *VoskRecognizer) Result() string
  • func (r *VoskRecognizer) SetEndpointerDelays(startMax, end, max float64)
  • func (r *VoskRecognizer) SetGrm(grammar string)
  • func (r *VoskRecognizer) SetMaxAlternatives(maxAlternatives int)
  • func (r *VoskRecognizer) SetPartialWords(words int)
  • func (r *VoskRecognizer) SetSpkModel(spkModel *VoskSpkModel)
  • func (r *VoskRecognizer) SetWords(words int)
• type VoskSpkModel
  • func NewSpkModel(spkModelPath string) (*VoskSpkModel, error)
  • func (s *VoskSpkModel) Free()

Functions

func GPUInit

func GPUInit()

GPUInit automatically selects a CUDA device and allows multithreading.

func GPUThreadInit

func GPUThreadInit()

GPUThreadInit inits CUDA device in a multi-threaded environment.

func SetLogLevel

func SetLogLevel(logLevel int)

SetLogLevel sets the log level for Kaldi messages.

Types

type VoskModel

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

VoskModel contains a reference to the C VoskModel

func NewModel

func NewModel(modelPath string) (*VoskModel, error)

NewModel creates a new VoskModel instance

func (*VoskModel) FindWord

func (m *VoskModel) FindWord(word string) int

FindWord checks if a word can be recognized by the model. Returns the word symbol if the word exists inside the model or -1 otherwise.

func (*VoskModel) Free

func (m *VoskModel) Free()

type VoskRecognizer

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

VoskRecognizer contains a reference to the C VoskRecognizer

func NewRecognizer

func NewRecognizer(model *VoskModel, sampleRate float64) (*VoskRecognizer, error)

NewRecognizer creates a new VoskRecognizer instance

func NewRecognizerGrm

func NewRecognizerGrm(model *VoskModel, sampleRate float64, grammar string) (*VoskRecognizer, error)

NewRecognizerGrm creates a new VoskRecognizer instance with the phrase list.

func NewRecognizerSpk

func NewRecognizerSpk(model *VoskModel, sampleRate float64, spkModel *VoskSpkModel) (*VoskRecognizer, error)

NewRecognizerSpk creates a new VoskRecognizer instance with a speaker model.

func (*VoskRecognizer) AcceptWaveform

func (r *VoskRecognizer) AcceptWaveform(buffer []byte) int

AcceptWaveform accepts and processes a new chunk of the voice data.

func (*VoskRecognizer) FinalResult

func (r *VoskRecognizer) FinalResult() string

FinalResult returns a speech recognition result. Same as result, but doesn't wait for silence.

func (*VoskRecognizer) Free

func (r *VoskRecognizer) Free()

func (*VoskRecognizer) PartialResult

func (r *VoskRecognizer) PartialResult() string

PartialResult returns a partial speech recognition result.

func (*VoskRecognizer) Reset

func (r *VoskRecognizer) Reset()

Reset resets the recognizer.

func (*VoskRecognizer) Result

func (r *VoskRecognizer) Result() string

Result returns a speech recognition result.

func (*VoskRecognizer) SetEndpointerDelays

func (r *VoskRecognizer) SetEndpointerDelays(startMax, end, max float64)

SetEndpointerDelays sets the recognition timeouts, where startMax is the timeout for stopping recognition in case of initial silence (usually around 5), end is the timeout for stopping recognition in milliseconds after we recognized something (usually around 0.5-1.0), and max is the timeout for forcing utterance end in milliseconds (usually around 20-30).

func (*VoskRecognizer) SetGrm

func (r *VoskRecognizer) SetGrm(grammar string)

SetGrm sets which phrases to recognize on an already initialized recognizer.

func (*VoskRecognizer) SetMaxAlternatives

func (r *VoskRecognizer) SetMaxAlternatives(maxAlternatives int)

SetMaxAlternatives configures the recognizer to output n-best results.

func (*VoskRecognizer) SetPartialWords

func (r *VoskRecognizer) SetPartialWords(words int)

SetPartialWords enables words with times in the partial ouput.

func (*VoskRecognizer) SetSpkModel

func (r *VoskRecognizer) SetSpkModel(spkModel *VoskSpkModel)

SetSpkModel adds a speaker model to an already initialized recognizer.

func (*VoskRecognizer) SetWords

func (r *VoskRecognizer) SetWords(words int)

SetWords enables words with times in the ouput.

type VoskSpkModel

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

VoskSpkModel contains a reference to the C VoskSpkModel

func NewSpkModel

func NewSpkModel(spkModelPath string) (*VoskSpkModel, error)

NewSpkModel creates a new VoskSpkModel instance

func (*VoskSpkModel) Free

func (s *VoskSpkModel) Free()