init

whisper.cpp-1.5.2/bindings/go/examples/go-model-download/context.go

@@ -0,0 +1,30 @@
+package main
+
+import (
+	"context"
+	"os"
+	"os/signal"
+)
+
+// ContextForSignal returns a context object which is cancelled when a signal
+// is received. It returns nil if no signal parameter is provided
+func ContextForSignal(signals ...os.Signal) context.Context {
+	if len(signals) == 0 {
+		return nil
+	}
+
+	ch := make(chan os.Signal)
+	ctx, cancel := context.WithCancel(context.Background())
+
+	// Send message on channel when signal received
+	signal.Notify(ch, signals...)
+
+	// When any signal received, call cancel
+	go func() {
+		<-ch
+		cancel()
+	}()
+
+	// Return success
+	return ctx
+}

whisper.cpp-1.5.2/bindings/go/examples/go-model-download/main.go

@@ -0,0 +1,208 @@
+package main
+
+import (
+	"context"
+	"flag"
+	"fmt"
+	"io"
+	"net/http"
+	"net/url"
+	"os"
+	"path/filepath"
+	"syscall"
+	"time"
+)
+
+///////////////////////////////////////////////////////////////////////////////
+// CONSTANTS
+
+const (
+	srcUrl  = "https://huggingface.co/ggerganov/whisper.cpp/resolve/main" // The location of the models
+	srcExt  = ".bin"                                                      // Filename extension
+	bufSize = 1024 * 64                                                   // Size of the buffer used for downloading the model
+)
+
+var (
+	// The models which will be downloaded, if no model is specified as an argument
+	modelNames = []string{"ggml-tiny.en", "ggml-tiny", "ggml-base.en", "ggml-base", "ggml-small.en", "ggml-small", "ggml-medium.en", "ggml-medium", "ggml-large-v1", "ggml-large-v2", "ggml-large-v3"}
+)
+
+var (
+	// The output folder. When not set, use current working directory.
+	flagOut = flag.String("out", "", "Output folder")
+
+	// HTTP timeout parameter - will timeout if takes longer than this to download a model
+	flagTimeout = flag.Duration("timeout", 30*time.Minute, "HTTP timeout")
+
+	// Quiet parameter - will not print progress if set
+	flagQuiet = flag.Bool("quiet", false, "Quiet mode")
+)
+
+///////////////////////////////////////////////////////////////////////////////
+// MAIN
+
+func main() {
+	flag.Usage = func() {
+		name := filepath.Base(flag.CommandLine.Name())
+		fmt.Fprintf(flag.CommandLine.Output(), "Usage: %s [options] <model>\n\n", name)
+		flag.PrintDefaults()
+	}
+	flag.Parse()
+
+	// Get output path
+	out, err := GetOut()
+	if err != nil {
+		fmt.Fprintln(os.Stderr, "Error:", err)
+		os.Exit(-1)
+	}
+
+	// Create context which quits on SIGINT or SIGQUIT
+	ctx := ContextForSignal(os.Interrupt, syscall.SIGQUIT)
+
+	// Progress filehandle
+	progress := os.Stdout
+	if *flagQuiet {
+		progress, err = os.Open(os.DevNull)
+		if err != nil {
+			fmt.Fprintln(os.Stderr, "Error:", err)
+			os.Exit(-1)
+		}
+		defer progress.Close()
+	}
+
+	// Download models - exit on error or interrupt
+	for _, model := range GetModels() {
+		url, err := URLForModel(model)
+		if err != nil {
+			fmt.Fprintln(os.Stderr, "Error:", err)
+			continue
+		} else if path, err := Download(ctx, progress, url, out); err == nil || err == io.EOF {
+			continue
+		} else if err == context.Canceled {
+			os.Remove(path)
+			fmt.Fprintln(progress, "\nInterrupted")
+			break
+		} else if err == context.DeadlineExceeded {
+			os.Remove(path)
+			fmt.Fprintln(progress, "Timeout downloading model")
+			continue
+		} else {
+			os.Remove(path)
+			fmt.Fprintln(os.Stderr, "Error:", err)
+			break
+		}
+	}
+}
+
+///////////////////////////////////////////////////////////////////////////////
+// PUBLIC METHODS
+
+// GetOut returns the path to the output directory
+func GetOut() (string, error) {
+	if *flagOut == "" {
+		return os.Getwd()
+	}
+	if info, err := os.Stat(*flagOut); err != nil {
+		return "", err
+	} else if !info.IsDir() {
+		return "", fmt.Errorf("not a directory: %s", info.Name())
+	} else {
+		return *flagOut, nil
+	}
+}
+
+// GetModels returns the list of models to download
+func GetModels() []string {
+	if flag.NArg() == 0 {
+		return modelNames
+	} else {
+		return flag.Args()
+	}
+}
+
+// URLForModel returns the URL for the given model on huggingface.co
+func URLForModel(model string) (string, error) {
+	if filepath.Ext(model) != srcExt {
+		model += srcExt
+	}
+	url, err := url.Parse(srcUrl)
+	if err != nil {
+		return "", err
+	} else {
+		url.Path = filepath.Join(url.Path, model)
+	}
+	return url.String(), nil
+}
+
+// Download downloads the model from the given URL to the given output directory
+func Download(ctx context.Context, p io.Writer, model, out string) (string, error) {
+	// Create HTTP client
+	client := http.Client{
+		Timeout: *flagTimeout,
+	}
+
+	// Initiate the download
+	req, err := http.NewRequest("GET", model, nil)
+	if err != nil {
+		return "", err
+	}
+	resp, err := client.Do(req)
+	if err != nil {
+		return "", err
+	}
+	defer resp.Body.Close()
+	if resp.StatusCode != http.StatusOK {
+		return "", fmt.Errorf("%s: %s", model, resp.Status)
+	}
+
+	// If output file exists and is the same size as the model, skip
+	path := filepath.Join(out, filepath.Base(model))
+	if info, err := os.Stat(path); err == nil && info.Size() == resp.ContentLength {
+		fmt.Fprintln(p, "Skipping", model, "as it already exists")
+		return "", nil
+	}
+
+	// Create file
+	w, err := os.Create(path)
+	if err != nil {
+		return "", err
+	}
+	defer w.Close()
+
+	// Report
+	fmt.Fprintln(p, "Downloading", model, "to", out)
+
+	// Progressively download the model
+	data := make([]byte, bufSize)
+	count, pct := int64(0), int64(0)
+	ticker := time.NewTicker(5 * time.Second)
+	for {
+		select {
+		case <-ctx.Done():
+			// Cancelled, return error
+			return path, ctx.Err()
+		case <-ticker.C:
+			pct = DownloadReport(p, pct, count, resp.ContentLength)
+		default:
+			// Read body
+			n, err := resp.Body.Read(data)
+			if err != nil {
+				DownloadReport(p, pct, count, resp.ContentLength)
+				return path, err
+			} else if m, err := w.Write(data[:n]); err != nil {
+				return path, err
+			} else {
+				count += int64(m)
+			}
+		}
+	}
+}
+
+// Report periodically reports the download progress when percentage changes
+func DownloadReport(w io.Writer, pct, count, total int64) int64 {
+	pct_ := count * 100 / total
+	if pct_ > pct {
+		fmt.Fprintf(w, "  ...%d MB written (%d%%)\n", count/1e6, pct_)
+	}
+	return pct_
+}

whisper.cpp-1.5.2/bindings/go/examples/go-whisper/color.go

@@ -0,0 +1,22 @@
+package main
+
+import "fmt"
+
+///////////////////////////////////////////////////////////////////////////////
+// CONSTANTS
+
+const (
+	Reset     = "\033[0m"
+	RGBPrefix = "\033[38;5;" // followed by RGB values in decimal format separated by colons
+	RGBSuffix = "m"
+)
+
+///////////////////////////////////////////////////////////////////////////////
+// PUBLIC METHODS
+
+// Colorize text with RGB values, from 0 to 23
+func Colorize(text string, v int) string {
+	// https://en.wikipedia.org/wiki/ANSI_escape_code#8-bit
+	// Grayscale colors are in the range 232-255
+	return RGBPrefix + fmt.Sprint(v%24+232) + RGBSuffix + text + Reset
+}

whisper.cpp-1.5.2/bindings/go/examples/go-whisper/flags.go

@@ -0,0 +1,156 @@
+package main
+
+import (
+	"flag"
+	"fmt"
+	"strings"
+	"time"
+
+	// Packages
+	whisper "github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
+)
+
+///////////////////////////////////////////////////////////////////////////////
+// TYPES
+
+type Flags struct {
+	*flag.FlagSet
+}
+
+///////////////////////////////////////////////////////////////////////////////
+// LIFECYCLE
+
+func NewFlags(name string, args []string) (*Flags, error) {
+	flags := &Flags{
+		FlagSet: flag.NewFlagSet(name, flag.ContinueOnError),
+	}
+
+	// Register the command line arguments
+	registerFlags(flags)
+
+	// Parse command line
+	if err := flags.Parse(args); err != nil {
+		return nil, err
+	}
+
+	// Return success
+	return flags, nil
+}
+
+///////////////////////////////////////////////////////////////////////////////
+// PUBLIC METHODS
+
+func (flags *Flags) GetModel() string {
+	return flags.Lookup("model").Value.String()
+}
+
+func (flags *Flags) GetLanguage() string {
+	return flags.Lookup("language").Value.String()
+}
+
+func (flags *Flags) IsTranslate() bool {
+	return flags.Lookup("translate").Value.(flag.Getter).Get().(bool)
+}
+
+func (flags *Flags) GetOffset() time.Duration {
+	return flags.Lookup("offset").Value.(flag.Getter).Get().(time.Duration)
+}
+
+func (flags *Flags) GetDuration() time.Duration {
+	return flags.Lookup("duration").Value.(flag.Getter).Get().(time.Duration)
+}
+
+func (flags *Flags) GetThreads() uint {
+	return flags.Lookup("threads").Value.(flag.Getter).Get().(uint)
+}
+
+func (flags *Flags) GetOut() string {
+	return strings.ToLower(flags.Lookup("out").Value.String())
+}
+
+func (flags *Flags) IsSpeedup() bool {
+	return flags.Lookup("speedup").Value.String() == "true"
+}
+
+func (flags *Flags) IsTokens() bool {
+	return flags.Lookup("tokens").Value.String() == "true"
+}
+
+func (flags *Flags) IsColorize() bool {
+	return flags.Lookup("colorize").Value.String() == "true"
+}
+
+func (flags *Flags) GetMaxLen() uint {
+	return flags.Lookup("max-len").Value.(flag.Getter).Get().(uint)
+}
+
+func (flags *Flags) GetMaxTokens() uint {
+	return flags.Lookup("max-tokens").Value.(flag.Getter).Get().(uint)
+}
+
+func (flags *Flags) GetWordThreshold() float32 {
+	return float32(flags.Lookup("word-thold").Value.(flag.Getter).Get().(float64))
+}
+
+func (flags *Flags) SetParams(context whisper.Context) error {
+	if lang := flags.GetLanguage(); lang != "" && lang != "auto" {
+		fmt.Fprintf(flags.Output(), "Setting language to %q\n", lang)
+		if err := context.SetLanguage(lang); err != nil {
+			return err
+		}
+	}
+	if flags.IsTranslate() && context.IsMultilingual() {
+		fmt.Fprintf(flags.Output(), "Setting translate to true\n")
+		context.SetTranslate(true)
+	}
+	if offset := flags.GetOffset(); offset != 0 {
+		fmt.Fprintf(flags.Output(), "Setting offset to %v\n", offset)
+		context.SetOffset(offset)
+	}
+	if duration := flags.GetDuration(); duration != 0 {
+		fmt.Fprintf(flags.Output(), "Setting duration to %v\n", duration)
+		context.SetDuration(duration)
+	}
+	if flags.IsSpeedup() {
+		fmt.Fprintf(flags.Output(), "Setting speedup to true\n")
+		context.SetSpeedup(true)
+	}
+	if threads := flags.GetThreads(); threads != 0 {
+		fmt.Fprintf(flags.Output(), "Setting threads to %d\n", threads)
+		context.SetThreads(threads)
+	}
+	if max_len := flags.GetMaxLen(); max_len != 0 {
+		fmt.Fprintf(flags.Output(), "Setting max_segment_length to %d\n", max_len)
+		context.SetMaxSegmentLength(max_len)
+	}
+	if max_tokens := flags.GetMaxTokens(); max_tokens != 0 {
+		fmt.Fprintf(flags.Output(), "Setting max_tokens to %d\n", max_tokens)
+		context.SetMaxTokensPerSegment(max_tokens)
+	}
+	if word_threshold := flags.GetWordThreshold(); word_threshold != 0 {
+		fmt.Fprintf(flags.Output(), "Setting word_threshold to %f\n", word_threshold)
+		context.SetTokenThreshold(word_threshold)
+	}
+
+	// Return success
+	return nil
+}
+
+///////////////////////////////////////////////////////////////////////////////
+// PRIVATE METHODS
+
+func registerFlags(flag *Flags) {
+	flag.String("model", "", "Path to the model file")
+	flag.String("language", "", "Spoken language")
+	flag.Bool("translate", false, "Translate from source language to english")
+	flag.Duration("offset", 0, "Time offset")
+	flag.Duration("duration", 0, "Duration of audio to process")
+	flag.Uint("threads", 0, "Number of threads to use")
+	flag.Bool("speedup", false, "Enable speedup")
+	flag.Uint("max-len", 0, "Maximum segment length in characters")
+	flag.Uint("max-tokens", 0, "Maximum tokens per segment")
+	flag.Float64("word-thold", 0, "Maximum segment score")
+	flag.Bool("tokens", false, "Display tokens")
+	flag.Bool("colorize", false, "Colorize tokens")
+	flag.String("out", "", "Output format (srt, none or leave as empty string)")
+}

whisper.cpp-1.5.2/bindings/go/examples/go-whisper/main.go

@@ -0,0 +1,43 @@
+package main
+
+import (
+	"flag"
+	"fmt"
+	"os"
+	"path/filepath"
+
+	// Packages
+	whisper "github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
+)
+
+func main() {
+	flags, err := NewFlags(filepath.Base(os.Args[0]), os.Args[1:])
+	if err == flag.ErrHelp {
+		os.Exit(0)
+	} else if err != nil {
+		fmt.Fprintln(os.Stderr, err)
+		os.Exit(1)
+	} else if flags.GetModel() == "" {
+		fmt.Fprintln(os.Stderr, "Use -model flag to specify which model file to use")
+		os.Exit(1)
+	} else if flags.NArg() == 0 {
+		fmt.Fprintln(os.Stderr, "No input files specified")
+		os.Exit(1)
+	}
+
+	// Load model
+	model, err := whisper.New(flags.GetModel())
+	if err != nil {
+		fmt.Fprintln(os.Stderr, err)
+		os.Exit(1)
+	}
+	defer model.Close()
+
+	// Process files
+	for _, filename := range flags.Args() {
+		if err := Process(model, filename, flags); err != nil {
+			fmt.Fprintln(os.Stderr, err)
+			continue
+		}
+	}
+}

whisper.cpp-1.5.2/bindings/go/examples/go-whisper/process.go

@@ -0,0 +1,132 @@
+package main
+
+import (
+	"fmt"
+	"io"
+	"os"
+	"time"
+
+	// Package imports
+	whisper "github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
+	wav "github.com/go-audio/wav"
+)
+
+func Process(model whisper.Model, path string, flags *Flags) error {
+	var data []float32
+
+	// Create processing context
+	context, err := model.NewContext()
+	if err != nil {
+		return err
+	}
+
+	// Set the parameters
+	if err := flags.SetParams(context); err != nil {
+		return err
+	}
+
+	fmt.Printf("\n%s\n", context.SystemInfo())
+
+	// Open the file
+	fmt.Fprintf(flags.Output(), "Loading %q\n", path)
+	fh, err := os.Open(path)
+	if err != nil {
+		return err
+	}
+	defer fh.Close()
+
+	// Decode the WAV file - load the full buffer
+	dec := wav.NewDecoder(fh)
+	if buf, err := dec.FullPCMBuffer(); err != nil {
+		return err
+	} else if dec.SampleRate != whisper.SampleRate {
+		return fmt.Errorf("unsupported sample rate: %d", dec.SampleRate)
+	} else if dec.NumChans != 1 {
+		return fmt.Errorf("unsupported number of channels: %d", dec.NumChans)
+	} else {
+		data = buf.AsFloat32Buffer().Data
+	}
+
+	// Segment callback when -tokens is specified
+	var cb whisper.SegmentCallback
+	if flags.IsTokens() {
+		cb = func(segment whisper.Segment) {
+			fmt.Fprintf(flags.Output(), "%02d [%6s->%6s] ", segment.Num, segment.Start.Truncate(time.Millisecond), segment.End.Truncate(time.Millisecond))
+			for _, token := range segment.Tokens {
+				if flags.IsColorize() && context.IsText(token) {
+					fmt.Fprint(flags.Output(), Colorize(token.Text, int(token.P*24.0)), " ")
+				} else {
+					fmt.Fprint(flags.Output(), token.Text, " ")
+				}
+			}
+			fmt.Fprintln(flags.Output(), "")
+			fmt.Fprintln(flags.Output(), "")
+		}
+	}
+
+	// Process the data
+	fmt.Fprintf(flags.Output(), "  ...processing %q\n", path)
+	context.ResetTimings()
+	if err := context.Process(data, cb, nil); err != nil {
+		return err
+	}
+
+	context.PrintTimings()
+
+	// Print out the results
+	switch {
+	case flags.GetOut() == "srt":
+		return OutputSRT(os.Stdout, context)
+	case flags.GetOut() == "none":
+		return nil
+	default:
+		return Output(os.Stdout, context, flags.IsColorize())
+	}
+}
+
+// Output text as SRT file
+func OutputSRT(w io.Writer, context whisper.Context) error {
+	n := 1
+	for {
+		segment, err := context.NextSegment()
+		if err == io.EOF {
+			return nil
+		} else if err != nil {
+			return err
+		}
+		fmt.Fprintln(w, n)
+		fmt.Fprintln(w, srtTimestamp(segment.Start), " --> ", srtTimestamp(segment.End))
+		fmt.Fprintln(w, segment.Text)
+		fmt.Fprintln(w, "")
+		n++
+	}
+}
+
+// Output text to terminal
+func Output(w io.Writer, context whisper.Context, colorize bool) error {
+	for {
+		segment, err := context.NextSegment()
+		if err == io.EOF {
+			return nil
+		} else if err != nil {
+			return err
+		}
+		fmt.Fprintf(w, "[%6s->%6s]", segment.Start.Truncate(time.Millisecond), segment.End.Truncate(time.Millisecond))
+		if colorize {
+			for _, token := range segment.Tokens {
+				if !context.IsText(token) {
+					continue
+				}
+				fmt.Fprint(w, " ", Colorize(token.Text, int(token.P*24.0)))
+			}
+			fmt.Fprint(w, "\n")
+		} else {
+			fmt.Fprintln(w, " ", segment.Text)
+		}
+	}
+}
+
+// Return srtTimestamp
+func srtTimestamp(t time.Duration) string {
+	return fmt.Sprintf("%02d:%02d:%02d,%03d", t/time.Hour, (t%time.Hour)/time.Minute, (t%time.Minute)/time.Second, (t%time.Second)/time.Millisecond)
+}