This commit is contained in:
Kar
2025-06-17 15:53:01 +05:30
commit 4d20931ecc
411 changed files with 180695 additions and 0 deletions

View File

@@ -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
}

View File

@@ -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)")
}

View File

@@ -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
}
}
}

View File

@@ -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)
}