whisper

Documentation

Overview

This is the higher-level speech-to-text whisper.cpp API for go

Index

• Constants
• Variables
• type Context
• type EncoderBeginCallback
• type Model
  • func New(path string) (Model, error)
• type ProgressCallback
• type Segment
• type SegmentCallback
• type Token

Constants

const SampleBits = whisper.SampleBits

SampleBits is the number of bytes per sample.

const SampleRate = whisper.SampleRate

SampleRate is the sample rate of the audio data.

Variables

var (
	ErrUnableToLoadModel    = errors.New("unable to load model")
	ErrInternalAppError     = errors.New("internal application error")
	ErrProcessingFailed     = errors.New("processing failed")
	ErrUnsupportedLanguage  = errors.New("unsupported language")
	ErrModelNotMultilingual = errors.New("model is not multilingual")
)

Types

type Context

type Context interface {
	SetLanguage(string) error // Set the language to use for speech recognition, use "auto" for auto detect language.
	SetTranslate(bool)        // Set translate flag
	IsMultilingual() bool     // Return true if the model is multilingual.
	Language() string         // Get language
	DetectedLanguage() string // Get detected language

	SetOffset(time.Duration)          // Set offset
	SetDuration(time.Duration)        // Set duration
	SetThreads(uint)                  // Set number of threads to use
	SetSplitOnWord(bool)              // Set split on word flag
	SetTokenThreshold(float32)        // Set timestamp token probability threshold
	SetTokenSumThreshold(float32)     // Set timestamp token sum probability threshold
	SetMaxSegmentLength(uint)         // Set max segment length in characters
	SetTokenTimestamps(bool)          // Set token timestamps flag
	SetMaxTokensPerSegment(uint)      // Set max tokens per segment (0 = no limit)
	SetAudioCtx(uint)                 // Set audio encoder context
	SetMaxContext(n int)              // Set maximum number of text context tokens to store
	SetBeamSize(n int)                // Set Beam Size
	SetEntropyThold(t float32)        // Set Entropy threshold
	SetInitialPrompt(prompt string)   // Set initial prompt
	SetTemperature(t float32)         // Set temperature
	SetTemperatureFallback(t float32) // Set temperature incrementation

	// Process mono audio data and return any errors.
	// If defined, newly generated segments are passed to the
	// callback function during processing.
	Process([]float32, EncoderBeginCallback, SegmentCallback, ProgressCallback) error

	// After process is called, return segments until the end of the stream
	// is reached, when io.EOF is returned.
	NextSegment() (Segment, error)

	IsBEG(Token) bool          // Test for "begin" token
	IsSOT(Token) bool          // Test for "start of transcription" token
	IsEOT(Token) bool          // Test for "end of transcription" token
	IsPREV(Token) bool         // Test for "start of prev" token
	IsSOLM(Token) bool         // Test for "start of lm" token
	IsNOT(Token) bool          // Test for "No timestamps" token
	IsLANG(Token, string) bool // Test for token associated with a specific language
	IsText(Token) bool         // Test for text token

	// Timings
	PrintTimings()
	ResetTimings()

	SystemInfo() string
}

Context is the speech recognition context.

type EncoderBeginCallback

type EncoderBeginCallback func() bool

EncoderBeginCallback is the callback function for checking if we want to continue processing. It is called during the Process function

type Model

type Model interface {
	io.Closer

	// Return a new speech-to-text context.
	NewContext() (Context, error)

	// Return true if the model is multilingual.
	IsMultilingual() bool

	// Return all languages supported.
	Languages() []string
}

Model is the interface to a whisper model. Create a new model with the function whisper.New(string)

func New

func New(path string) (Model, error)

type ProgressCallback

type ProgressCallback func(int)

ProgressCallback is the callback function for reporting progress during processing. It is called during the Process function

type Segment

type Segment struct {
	// Segment Number
	Num int

	// Time beginning and end timestamps for the segment.
	Start, End time.Duration

	// The text of the segment.
	Text string

	// The tokens of the segment.
	Tokens []Token
}

Segment is the text result of a speech recognition.

type SegmentCallback

type SegmentCallback func(Segment)

SegmentCallback is the callback function for processing segments in real time. It is called during the Process function

type Token

type Token struct {
	Id         int
	Text       string
	P          float32
	Start, End time.Duration
}

Token is a text or special token