Using a speech engine to highlight words
Display a piece of text and produce spoken output via a speech engine.
As each word is being spoken, highlight the word on the display.
In languages where cursor control and highlighting are not possible, it is permissible to output each word as it is spoken.
- Related task
AutoHotkey
We use the simple SAPI.SPVoice COM Object and a parsing loop. The highlighting is done with EM_SETSEL and Notepad. Rather crude, but it works. Due to the simplistic nature of the parsing loop, the text ends with a space. <lang AutoHotkey>SetTitleMatchMode 2 EM_SETSEL := 0x00B1
Run notepad,,,pid WinWaitActive ahk_pid %pid% ControlSetText, Edit1, % text := "AutoHotkey was the first to implement this task! ", ahk_pid %pid%
pVoice := ComObjCreate("Sapi.spvoice"), i := 1 ; the spvoice COM Object ships with the OS
- parse the text
While lf := SubStr(text, i, 1) {
If lf = %A_Space%
{
SendMessage, EM_SetSel, % i-StrLen(word)-1, % i-1, Edit1, ahk_pid %pid%
pVoice.speak(word), word := "", i++
}
Else word .= lf, i++
}</lang>
Go
This uses the eSpeak speech synthesizer which is invoked for each word in the text. As the word is spoken it is printed to the terminal in capitalized form (and the previous word is uncapitalized). After a second's delay the final word is uncapitalized.
Very robotic but it works. <lang go>package main
import (
"fmt" "log" "os/exec" "strings" "time"
)
func main() {
s := "Actions speak louder than words."
prev := ""
prevLen := 0
bs := ""
for _, word := range strings.Fields(s) {
cmd := exec.Command("espeak", word)
if err := cmd.Run(); err != nil {
log.Fatal(err)
}
if prevLen > 0 {
bs = strings.Repeat("\b", prevLen)
}
fmt.Printf("%s%s%s ", bs, prev, strings.ToUpper(word))
prev = word + " "
prevLen = len(word) + 1
}
bs = strings.Repeat("\b", prevLen)
time.Sleep(time.Second)
fmt.Printf("%s%s\n", bs, prev)
}</lang>
Julia
<lang julia>function speak(sentence, cmd = "/utl/espeak.bat")
for word in split(sentence)
s = replace(lowercase(word), r"[^a-z]" => "")
if length(s) > 0
print(uppercase(s))
run(`$cmd $s`)
sleep(1)
print("\b"^length(s))
end
print(word, " ")
end
end
speak("Are those 144 shy Eurasian footwear, cowboy chaps, or jolly earthmoving headgear?") </lang>
M2000 Interpreter
<lang M2000 Interpreter> Module UsingEvents {
Form 60, 32
Cls 5, 0
Pen 14
Declare WithEvents sp "SAPI.SpVoice"
That$="Rosetta Code is a programming chrestomathy site"
margin=(width-Len(That$))/2
EndStream=False
\\ this function called as sub routine - same scope as Module
\\ we can call it from event function too
Function Localtxt {
\\ move the cursor to middle line
Cursor 0, height/2
\\ using OVER the line erased with background color and then print text over
\\ ordinary Print using transparent printing of text
\\ $(0) set mode to non proportional text, @() move the cursor to sepecific position
Print Over $(0),@(margin), That$
}
Call Local LocalTxt()
Function sp_Word {
Read New StreamNumber, StreamPosition, CharacterPosition, Length
Call Local LocalTxt()
Cursor 0, height/2
Pen 15 {Print Part $(0), @(CharacterPosition+margin); Mid$(That$, CharacterPosition+1, Length)}
Refresh
}
Function sp_EndStream {
Refresh
EndStream=True
}
Const SVEEndInputStream = 4
Const SVEWordBoundary = 32
Const SVSFlagsAsync = 1&
With sp, "EventInterests", SVEWordBoundary+SVEEndInputStream
Method sp, "Speak", That$, SVSFlagsAsync
While Not EndStream {Wait 10}
Call Local LocalTxt()
} UsingEvents
</lang>
Mathematica
<lang>DynamicModule[{text = "This is some text.", words, i = 0},
Panel@Column@{Dynamic[
Row[Riffle[
If[i != 0, MapAt[Style[#, Red] &, #, i], #] &@(words =
StringSplit@text), " "]]], InputField[Dynamic@text, String],
Button["Speak",
While[i < Length@words, i++; FinishDynamic[]; Speak[wordsi];
Pause[Max[0.7, 0.12 StringLength[wordsi]]]]; i = 0]}]</lang>
Nim
Works on Linux but may also work on other platforms provided "espeak" is installed. <lang Nim>import os, osproc, strutils
const S = "Actions speak louder than words."
var prev, bs = "" var prevlen = 0
for word in S.splitWhitespace():
discard execProcess("espeak " & word)
if prevlen > 0:
bs = repeat('\b', prevlen)
stdout.write bs, prev, word.toUpper, ' '
stdout.flushFile()
prev = word & ' '
prevlen = word.len + 1
bs = repeat('\b', prevlen) sleep(1000) echo bs, prev</lang>
Phix
<lang Phix>-- demo/rosetta/Speech.exw requires(WINDOWS) requires(32) -- Windows 32 bit only, for now... (runs fine on a 64-bit OS, but needs a 32-bit pw.exe) include pGUI.e Ihandle t3, left, red, right, btn, bc, dlg
include pComN.ew
procedure speak(atom pVoice, string text, integer flags=NULL)
{} = cominvk(pVoice,ISpVoice_Speak,{unicode_string(text),flags,0})
end procedure
procedure set_interest(atom pVoice, sequence flags)
atom lo = #40000000,
hi = #00000002
for i=1 to length(flags) do
lo += power(2,flags[i])
end for
-- NB 2*64-bit as 4*32-bit:
atom res = cominvk(pVoice,ISpVoice_SetInterest,{lo,hi,lo,hi})
if res!=S_OK then crash("invalid") end if
end procedure
constant text = `<rate speed="-5">Highlight words as they are spoken.` atom pVoice = NULL
include cffi.e constant tSPEVENT = """ typedef struct SPEVENT {
WORD eEventId; WORD elParamType; ULONG ulStreamNum; ULONGLONG ullAudioStreamOffset; WPARAM wParam; LPARAM lParam;
} SPEVENT;""", idSPEVENT = define_struct(tSPEVENT), pSPEVENT = allocate_struct(idSPEVENT)
function eventHandler(atom wParam, lParam)
while cominvk(pVoice,ISpVoice_GetEvents,{1,pSPEVENT,NULL})=S_OK do
integer eEventId = get_struct_field(idSPEVENT,pSPEVENT,"eEventId")
if eEventId=SPEI_WORD_BOUNDARY then
integer len = get_struct_field(idSPEVENT,pSPEVENT,"wParam"),
pos = get_struct_field(idSPEVENT,pSPEVENT,"lParam")
IupSetStrAttribute(left,"TITLE",text[1..pos])
IupSetStrAttribute(red,"TITLE",text[pos+1..pos+len])
IupSetStrAttribute(right,"TITLE",text[pos+len+1..$])
elsif eEventId=SPEI_END_INPUT_STREAM then
IupSetStrAttribute(left,"TITLE",text)
IupSetStrAttribute(red,"TITLE","")
IupSetStrAttribute(right,"TITLE","")
end if
IupSetAttribute(left, "RASTERSIZE", "x0")
IupSetAttribute(red, "RASTERSIZE", "x0")
IupSetAttribute(right, "RASTERSIZE", "x0")
IupRefresh(t3)
IupRedraw(t3)
end while
return 0
end function constant cb_eventHandler = call_back(eventHandler)
function button_cb(Ihandle /*ih*/)
object res
if platform()=WINDOWS then
if pVoice=NULL then
pVoice = allocate(machine_word())
res = CoCreateInstance(CLSID_SpVoice,IID_ISpVoice,pVoice)
if res!=S_OK then
crash("Failed to initialize SpeechAPI. (%08x)\n",res)
end if
end if
if cominvk(pVoice,ISpVoice_SetNotifyCallbackFunction,{cb_eventHandler,0,0})!=S_OK then ?9/0 end if
set_interest(pVoice, {SPEI_WORD_BOUNDARY,SPEI_END_INPUT_STREAM})
speak(pVoice,text,SPF_IS_XML+SPF_ASYNC)
else
-- DEV no highlighting here...
{} = system_exec(sprintf(`espeak "%s"`,{text}))
end if
return IUP_CONTINUE
end function
procedure main()
IupOpen()
left = IupLabel("",`FONT="Verdana, 18"`)
red = IupLabel("",`FONT="Verdana, 18", FGCOLOR="255 0 0"`)
right = IupLabel(text,`FONT="Verdana, 18"`)
t3 = IupHbox({IupFill(),left,red,right,IupFill()},"MARGIN=20x20")
btn = IupButton("Speak",Icallback("button_cb"))
bc = IupHbox({IupFill(),btn,IupFill()},"MARGIN=10x10")
dlg = IupDialog(IupVbox({t3,bc}),"TITLE=Speak")
IupShow(dlg)
IupMainLoop()
IupClose()
end procedure
main()</lang>
Raku
<lang perl6># 20200826 Raku programming solution
my \s = "Actions speak louder than words."; my \prev = $ = ""; my \prevLen = $ = 0; my \bs = $ = "";
for s.split(' ', :skip-empty) {
run "/usr/bin/espeak", $_ or die; bs = "\b" x prevLen if prevLen > 0; printf "%s%s%s ", bs, prev, $_.uc; prev = "$_ "; prevLen = $_.chars + 1
}
printf "%s%s\n", "\b" x prevLen, prev;</lang>
REXX
Programming note: This REXX program uses a freeware program NIRCMD to interface with the Microsoft Windows speech synthesizer program SAM, a text to speech using a male voice. SAM can possibly be configured to use other voices with later releases of Windows. More recent Microsoft Windows have another speech synthesizer program: ANNA.
Each word of the text is highlighted (by showing the word in uppercase). the terminal screen is cleared before showing the text that is being spoken; the repeated calls to the (Windows) speech engine makes for a slower speech rate. <lang rexx>/*REXX program uses a command line interface to invoke Windows SAM for speech synthesis.*/ parse arg t /*get the (optional) text from the C.L.*/
- = words(t)
if #==0 then exit /*Nothing to say? Then exit program.*/ dq= '"' /*needed to enclose text in dbl quotes.*/ rate= 1 /*talk: -10 (slow) to 10 (fast). */
/* [↓] where the rubber meets the road*/
do j=1 for #
x= word(t, j); upper x /*extract 1 word, capitalize it for HL.*/
if j==1 then LHS= /*obtain text before the spoken word. */
else LHS= subword(t, 1, j-1)
if j==# then RHS= /*obtain text after the spoken word. */
else RHS= subword(t, j+1)
'CLS' /*use this command to clear the screen.*/
say 'speaking: ' space(LHS x RHS) /*show text, one word is capitalized. */
oneWord= dq x dq /*surround a word in double quotes (").*/
'NIRCMD' "speak text" oneWord rate /*NIRCMD invokes Microsoft's Sam voice*/
end /*j*/ /*stick a fork in it, we're all done. */</lang>
Note: The name of the above REXX program is SPEAKHI.REX
usage using the command:
speakhi This is an example of speech synthesis.
Ring
<lang ring> load "guilib.ring"
MyApp = New qApp {
win1 = new qWidget() {
setwindowtitle("Hello World")
setGeometry(100,100,370,250)
Text = "Welcome to the Ring Programming Language"
Text = split(Text," ")
label1 = new qLabel(win1) {
settext("What is your name ?")
setGeometry(10,20,350,30)
setalignment(Qt_AlignHCenter)
}
btn1 = new qpushbutton(win1) {
setGeometry(10,200,100,30)
settext("Say Hello")
setclickevent("pHello()")
}
btn2 = new qpushbutton(win1) {
setGeometry(150,200,100,30)
settext("Close")
setclickevent("pClose()")
}
lineedit1 = new qlineedit(win1) {
setGeometry(10,100,350,30)
}
voice = new QTextToSpeech(win1) {
}
show()
}
exec()}
Func pHello
lineedit1.settext( "Hello " + lineedit1.text())
for n = 1 to len(Text)
voice.Say(Text[n])
see Text[n] + nl
next
Func pClose
MyApp.quit()
</lang>
- Output:
Welcome to the Ring Programming Language
Ruby
I'm having difficulty figuring out how to get Shoes to update the GUI (like Tk's update command), so the user must click the button once for each word.
Uses the Ruby code from Speech synthesis <lang ruby>load 'speechsynthesis.rb'
if ARGV.length == 1
$text = "This is default text for the highlight and speak program"
else
$text = ARGV[1..-1].join(" ")
end $words = $text.split
Shoes.app do
@idx = 0
stack do
@sentence = para(strong($words[0] + " "), $words[1..-1].map {|word| span(word + " ")})
button "Say word" do
say_and_highlight
end
end
keypress do |key| case key when :control_q, "\x11" then exit end end
def say_and_highlight
speak $words[@idx]
@idx = (@idx + 1) % $words.length
@sentence.replace($words.each_with_index.map {|word, idx| idx == @idx ? strong(word + " ") : span(word + " ")})
end
end</lang>
Tcl
This code uses the external /usr/bin/say program (known available on Mac OS X) as its interface to the speech engine; this produces rather stilted speech because it forces the text to be spoken one word at a time instead of as a whole sentence (in order to keep the highlighting synchronized).
<lang tcl>package require Tcl 8.5 package require Tk 8.5 proc say {text button} {
grab $button
$button configure -state disabled -cursor watch
update
set starts [$text search -all -regexp -count lengths {\S+} 1.0]
foreach start $starts length $lengths {
lappend strings [$text get $start "$start + $length char"] lappend ends [$text index "$start + $length char"]
}
$text tag remove sel 1.0 end
foreach from $starts str $strings to $ends {
$text tag add sel $from $to update idletasks exec /usr/bin/say << $str $text tag remove sel 1.0 end
}
grab release $button
$button configure -state normal -cursor {}
}
pack [text .t] pack [button .b -text "Speak, computer!" -command {say .t .b}] -fill x .t insert 1.0 "This is an example of speech synthesis with Tcl/Tk."</lang>