main.go

// +build windows
package main

import (
	"flag"
	"fmt"
	"io/ioutil"
	"log"
	"os"
	"reflect"
	"strings"
	"syscall"
	"unsafe"

	"github.com/go-ole/go-ole"
	"github.com/moutend/go-wss"
)

type FilenameFlag struct {
	Value string
}

func (f *FilenameFlag) Set(value string) (err error) {
	if !strings.HasSuffix(value, ".wav") {
		err = fmt.Errorf("specify WAVE audio file (*.wav)")
		return
	}
	f.Value = value
	return
}

func (f *FilenameFlag) String() string {
	return f.Value
}

func GetSpeechStreamSync(text string) (stream []byte, err error) {
	var modSpeechSynthesizer, _ = syscall.LoadDLL("speechsynthesizer.dll")
	var procGetSpeechStreamSync, _ = modSpeechSynthesizer.FindProc("CreateSpeechStream")
	var buf *wss.IBuffer

	ole.RoInitialize(1)
	//defer ole.RoUninitialize()

	var textHString ole.HString
	if text == "" {
		text = "Hello, World!"
	}
	if textHString, err = ole.NewHString(text); err != nil {
		return
	}
	defer ole.DeleteHString(textHString)

	hr, _, _ := procGetSpeechStreamSync.Call(
		uintptr(unsafe.Pointer(textHString)),
		uintptr(unsafe.Pointer(&buf)))
	if hr != 0 {
		err = fmt.Errorf("unknown error %v", hr)
		return
	}
	defer buf.Release()
	var length uint32
	if err = buf.GetLength(&length); err != nil {
		return
	}

	// Casting Ibuffer to IBufferByteAccess
	var unk *ole.IUnknown
	if buf.PutQueryInterface(ole.IID_IUnknown, &unk); err != nil {
		return
	}

	var bba *wss.IBufferByteAccess
	if err = unk.PutQueryInterface(wss.IID_IBufferByteAccess, &bba); err != nil {
		return
	}

	// Extract the backing array.
	var bufPtr *byte
	if err = bba.Buffer(&bufPtr); err != nil {
		return
	}

	rawBufPtr := uintptr(unsafe.Pointer(bufPtr))
	sliceHeader := reflect.SliceHeader{Data: rawBufPtr, Len: int(length), Cap: int(length)}
	stream = *(*[]byte)(unsafe.Pointer(&sliceHeader))

	return
}

func main() {
	var err error
	if err = run(os.Args); err != nil {
		log.Fatal(err)
	}
	return
}

func run(args []string) (err error) {
	var filenameFlag FilenameFlag
	var stream []byte
	var text string

	f := flag.NewFlagSet(args[0], flag.ExitOnError)
	f.Var(&filenameFlag, "output", "file name")
	f.Var(&filenameFlag, "o", "Alias of --output")
	f.Parse(args[1:])

	text = strings.Join(f.Args(), " ")
	if filenameFlag.Value == "" {
		filenameFlag.Value = "voice.wav"
	}
	if stream, err = GetSpeechStreamSync(text); err != nil {
		return
	}
	return ioutil.WriteFile(filenameFlag.Value, stream, 0644)
}