Preserve prosody-relevant punctuation in espeakbridge

libpiper/include/piper_impl.hpp

@@ -45,6 +45,10 @@ Ort::Env ort_env{ORT_LOGGING_LEVEL_WARNING, "piper"};
 
 #define CLAUSE_TYPE_CLAUSE 0x00040000
 #define CLAUSE_TYPE_SENTENCE 0x00080000
+#define CLAUSE_TYPE_EOF 0x00010000
+
+#define CLAUSE_INTONATION_TYPE 0x00007000
+#define CLAUSE_PAUSE 0x00000FFF
 
 #define CLAUSE_PERIOD (40 | CLAUSE_INTONATION_FULL_STOP | CLAUSE_TYPE_SENTENCE)
 #define CLAUSE_COMMA (20 | CLAUSE_INTONATION_COMMA | CLAUSE_TYPE_CLAUSE)
@@ -54,6 +58,11 @@ Ort::Env ort_env{ORT_LOGGING_LEVEL_WARNING, "piper"};
 #define CLAUSE_COLON (30 | CLAUSE_INTONATION_FULL_STOP | CLAUSE_TYPE_CLAUSE)
 #define CLAUSE_SEMICOLON (30 | CLAUSE_INTONATION_COMMA | CLAUSE_TYPE_CLAUSE)
 
+#define EXCLAMATION_PAUSE (CLAUSE_EXCLAMATION & CLAUSE_PAUSE)
+#define PERIOD_PAUSE (CLAUSE_PERIOD & CLAUSE_PAUSE)
+#define SEMICOLON_PAUSE (CLAUSE_SEMICOLON & CLAUSE_PAUSE)
+#define COMMA_PAUSE (CLAUSE_COMMA & CLAUSE_PAUSE)
+
 struct piper_synthesizer {
     // From config JSON file
     std::string espeak_voice;

libpiper/src/piper.cpp

@@ -129,6 +129,29 @@ piper_default_synthesize_options(piper_synthesizer *synth) {
     return options;
 }
 
+std::string categorize_terminator(int terminator) {
+    if ((terminator & CLAUSE_TYPE_EOF) != CLAUSE_TYPE_EOF) {
+        int pause = terminator & CLAUSE_PAUSE;
+
+        if (pause >= PERIOD_PAUSE) {
+            switch (terminator & CLAUSE_INTONATION_TYPE) {
+                case CLAUSE_INTONATION_EXCLAMATION: return "!";
+                case CLAUSE_INTONATION_QUESTION: return "?";
+                default: return ".";
+            }
+        } else if (pause >= SEMICOLON_PAUSE) {
+            switch (terminator & CLAUSE_INTONATION_TYPE) {
+                case CLAUSE_INTONATION_FULL_STOP: return ":";
+                default: return ";";
+            }
+        } else if (pause >= COMMA_PAUSE) {
+            return ",";
+        }
+    }
+
+    return "";
+}
+
 int piper_synthesize_start(struct piper_synthesizer *synth, const char *text,
                            const piper_synthesize_options *options) {
     if (!synth) {
@@ -163,7 +186,6 @@ int piper_synthesize_start(struct piper_synthesizer *synth, const char *text,
     const void *text_ptr = text;
     while (text_ptr != nullptr) {
         int terminator = 0;
-        std::string terminator_str = "";
 
         const char *phonemes = espeak_TextToPhonemesWithTerminator(
             &text_ptr, espeakCHARS_AUTO, espeakPHONEMES_IPA, &terminator);
@@ -172,22 +194,7 @@ int piper_synthesize_start(struct piper_synthesizer *synth, const char *text,
             sentence_phonemes[current_idx] += phonemes;
         }
 
-        // Categorize terminator
-        terminator &= 0x000FFFFF;
-
-        if (terminator == CLAUSE_PERIOD) {
-            terminator_str = ".";
-        } else if (terminator == CLAUSE_QUESTION) {
-            terminator_str = "?";
-        } else if (terminator == CLAUSE_EXCLAMATION) {
-            terminator_str = "!";
-        } else if (terminator == CLAUSE_COMMA) {
-            terminator_str = ", ";
-        } else if (terminator == CLAUSE_COLON) {
-            terminator_str = ": ";
-        } else if (terminator == CLAUSE_SEMICOLON) {
-            terminator_str = "; ";
-        }
+        std::string terminator_str = categorize_terminator(terminator);
 
         sentence_phonemes[current_idx] += terminator_str;
 

pyproject.toml

@@ -8,3 +8,6 @@ exclude = '''
     \.pyi$
 )
 '''
+
+[tool.pytest.ini_options]
+disable_test_id_escaping_and_forfeit_all_rights_to_community_support = true

src/piper/espeakbridge.c

@@ -11,6 +11,10 @@
 
 #define CLAUSE_TYPE_CLAUSE 0x00040000
 #define CLAUSE_TYPE_SENTENCE 0x00080000
+#define CLAUSE_TYPE_EOF 0x00010000
+
+#define CLAUSE_INTONATION_TYPE 0x00007000
+#define CLAUSE_PAUSE 0x00000FFF
 
 #define CLAUSE_PERIOD (40 | CLAUSE_INTONATION_FULL_STOP | CLAUSE_TYPE_SENTENCE)
 #define CLAUSE_COMMA (20 | CLAUSE_INTONATION_COMMA | CLAUSE_TYPE_CLAUSE)
@@ -20,6 +24,11 @@
 #define CLAUSE_COLON (30 | CLAUSE_INTONATION_FULL_STOP | CLAUSE_TYPE_CLAUSE)
 #define CLAUSE_SEMICOLON (30 | CLAUSE_INTONATION_COMMA | CLAUSE_TYPE_CLAUSE)
 
+#define EXCLAMATION_PAUSE (CLAUSE_EXCLAMATION & CLAUSE_PAUSE)
+#define PERIOD_PAUSE (CLAUSE_PERIOD & CLAUSE_PAUSE)
+#define SEMICOLON_PAUSE (CLAUSE_SEMICOLON & CLAUSE_PAUSE)
+#define COMMA_PAUSE (CLAUSE_COMMA & CLAUSE_PAUSE)
+
 static PyObject *py_initialize(PyObject *self, PyObject *args) {
     const char *data_dir;
     if (!PyArg_ParseTuple(args, "s", &data_dir)) {
@@ -48,6 +57,29 @@ static PyObject *py_set_voice(PyObject *self, PyObject *args) {
     Py_RETURN_NONE;
 }
 
+static char *categorize_terminator(int terminator) {
+    if ((terminator & CLAUSE_TYPE_EOF) != CLAUSE_TYPE_EOF) {
+        int pause = terminator & CLAUSE_PAUSE;
+
+        if (pause >= PERIOD_PAUSE) {
+            switch (terminator & CLAUSE_INTONATION_TYPE) {
+                case CLAUSE_INTONATION_EXCLAMATION: return "!";
+                case CLAUSE_INTONATION_QUESTION: return "?";
+                default: return ".";
+            }
+        } else if (pause >= SEMICOLON_PAUSE) {
+            switch (terminator & CLAUSE_INTONATION_TYPE) {
+                case CLAUSE_INTONATION_FULL_STOP: return ":";
+                default: return ";";
+            }
+        } else if (pause >= COMMA_PAUSE) {
+            return ",";
+        }
+    }
+
+    return "";
+}
+
 static PyObject *py_get_phonemes(PyObject *self, PyObject *args) {
     const char *text;
     if (!PyArg_ParseTuple(args, "s", &text)) {
@@ -58,28 +90,12 @@ static PyObject *py_get_phonemes(PyObject *self, PyObject *args) {
 
     while (text != NULL) {
         int terminator = 0;
-        char *terminator_str = "";
 
         const char *phonemes = espeak_TextToPhonemesWithTerminator(
             (const void **)&text, espeakCHARS_AUTO, espeakPHONEMES_IPA,
             &terminator);
 
-        // Categorize terminator
-        terminator &= 0x000FFFFF;
-
-        if (terminator == CLAUSE_PERIOD) {
-            terminator_str = ".";
-        } else if (terminator == CLAUSE_QUESTION) {
-            terminator_str = "?";
-        } else if (terminator == CLAUSE_EXCLAMATION) {
-            terminator_str = "!";
-        } else if (terminator == CLAUSE_COMMA) {
-            terminator_str = ",";
-        } else if (terminator == CLAUSE_COLON) {
-            terminator_str = ":";
-        } else if (terminator == CLAUSE_SEMICOLON) {
-            terminator_str = ";";
-        }
+        char *terminator_str = categorize_terminator(terminator);
 
         PyList_Append(phonemes_and_terminators,
                       Py_BuildValue("(ssO)", phonemes, terminator_str,

tests/test_espeakbridge.py

@@ -0,0 +1,24 @@
+import pytest
+
+from piper import espeakbridge
+
+terminator_tests = [
+    # exact matches for punctuation are retained verbatim
+    ("en-us", "Hello, world! Colons: important? Maybe; maybe not.", ",!:?;."),
+    # ellipses get normalized to semicolons
+    ("en-us", "An ellipsis… Could it be true?", ";?"),
+    ("en-us", "An ASCII ellipsis... Could it be true?", ";?"),
+    ("en-us", "A long ellipsis.... Could it be true?", ";?"),
+    # em dashes get normalized to semicolons
+    ("en-us", "Em dashes — for all your punctuation needs.", ";."),
+    # Chinese/fullwidth punctuation gets normalized to ASCII equivalents
+    ("cmn", "你好，世界！冒号：重要？可能；与否。一、二、三……", ",!:?;.,,;"),
+]
+
+
+@pytest.mark.parametrize("locale,text,expect", terminator_tests)
+def test_get_phonemes_terminator(locale: str, text: str, expect: list[str]):
+    espeakbridge.set_voice(locale)
+    result = espeakbridge.get_phonemes(text)
+
+    assert [terminator for _, terminator, *_ in result] == list(expect)