ceos/espeakup
1
0
Fork 0
mirror of https://github.com/linux-speakup/espeakup.git synced 2026-08-09 09:19:09 +02:00
Code Issues Projects Releases Packages Wiki Activity

espeak: escape XML-special characters in single-character mode

Browse source 
When speakup echoes a single character, espeakup wraps it in SSML:

    <say-as interpret-as="characters">%c</say-as>

If the character is one of XML's five special characters (< > & ' "),
the raw byte is injected straight into the markup, producing ill-formed
SSML.  espeak-ng then misparses the <say-as> element, which surfaces as
a spurious high-pitched "ringing" whenever one of these characters is
spoken.

Escape each special character to its corresponding XML entity
(&lt; &gt; &amp; &apos; &quot;) so the markup stays well-formed.
Non-printable characters (< 0x20 or > 0x7e) are no longer wrapped in
SSML either; they fall through to the existing raw-synthesis fallback
so they cannot corrupt the surrounding element either.

The space and ordinary-printable-character paths are unchanged.
This commit is contained in:
donovanmalisch23-alt 2026-08-03 11:04:45 +00:00 • committed by Samuel Thibault
commit f266faaff7
1 changed files with 30 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

36
src/espeak.c
View file

@ -302,15 +302,39 @@ static espeak_ERROR speak_text(struct synth_t *s)
synth_mode |= espeakSSML;
if (espeakup_mode == ESPEAKUP_MODE_SPEAKUP && (s->len == 1)) {
char *buf;
char *buf = NULL;
int n;
if (s->buf[0] == ' ')
unsigned char c = s->buf[0];
if (c == ' ')
n = asprintf(&buf,
"<say-as interpret-as=\"tts:char\">&#32;</say-as>");
else
n = asprintf(&buf,
"<say-as interpret-as=\"characters\">%c</say-as>",
s->buf[0]);
else if (c < 0x20 || c > 0x7e) {
/* Not a printable character; do not embed it in SSML, as
* that would produce invalid markup. Fall through to the
* raw-synthesis path below. */
n = -1;
} else {
/* Escape characters that are special in XML/SSML so the
* resulting markup stays well-formed; otherwise espeak-ng
* misparses the <say-as> element, which is the source of
* the spurious high-pitched "ringing" on these characters. */
const char *entity = NULL;
switch (c) {
case '<': entity = "&lt;"; break;
case '>': entity = "&gt;"; break;
case '&': entity = "&amp;"; break;
case '\'': entity = "&apos;"; break;
case '"': entity = "&quot;"; break;
}
if (entity)
n = asprintf(&buf,
"<say-as interpret-as=\"characters\">%s</say-as>",
entity);
else
n = asprintf(&buf,
"<say-as interpret-as=\"characters\">%c</say-as>",
c);
}
if (n == -1) {
/* D'oh. Not much to do on allocation failure.
* Perhaps espeak will happen to say the character */