Skip to content

Fix punctuations in kokoro tts. #2458

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Aug 7, 2025
Merged

Fix punctuations in kokoro tts. #2458

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 5 additions & 2 deletions sherpa-onnx/csrc/kokoro-multi-lang-lexicon.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,7 +69,9 @@ class KokoroMultiLangLexicon::Impl {

std::vector<TokenIDs> ConvertTextToTokenIds(const std::string &_text,
const std::string &voice) const {
std::string text = ToLowerCase(_text);
// we cannot convert text to lowercase here since it will affect
// how piper_phonemize handles punctuations inside the text
std::string text = _text;
if (debug_) {
SHERPA_ONNX_LOGE("After converting to lowercase:\n%s", text.c_str());
}
Expand Down Expand Up @@ -300,7 +302,8 @@ class KokoroMultiLangLexicon::Impl {

this_sentence.push_back(0);

for (const auto &word : words) {
for (const auto &_word : words) {
auto word = ToLowerCase(_word);
if (IsPunctuation(word)) {
this_sentence.push_back(token2id_.at(word));

Expand Down
30 changes: 30 additions & 0 deletions sherpa-onnx/csrc/piper-phonemize-lexicon.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,6 +32,32 @@

namespace sherpa_onnx {

// Encode a single char32_t to UTF-8 string. For debugging only
static std::string ToString(char32_t cp) {
std::string result;

if (cp <= 0x7F) {
result += static_cast<char>(cp);
} else if (cp <= 0x7FF) {
result += static_cast<char>(0xC0 | ((cp >> 6) & 0x1F));
result += static_cast<char>(0x80 | (cp & 0x3F));
} else if (cp <= 0xFFFF) {
result += static_cast<char>(0xE0 | ((cp >> 12) & 0x0F));
result += static_cast<char>(0x80 | ((cp >> 6) & 0x3F));
result += static_cast<char>(0x80 | (cp & 0x3F));
} else if (cp <= 0x10FFFF) {
result += static_cast<char>(0xF0 | ((cp >> 18) & 0x07));
result += static_cast<char>(0x80 | ((cp >> 12) & 0x3F));
result += static_cast<char>(0x80 | ((cp >> 6) & 0x3F));
result += static_cast<char>(0x80 | (cp & 0x3F));
} else {
SHERPA_ONNX_LOGE("Invalid Unicode code point: %d",
static_cast<int32_t>(cp));
}

return result;
}

void CallPhonemizeEspeak(const std::string &text,
piper::eSpeakPhonemeConfig &config, // NOLINT
std::vector<std::vector<piper::Phoneme>> *phonemes) {
Expand Down Expand Up @@ -165,6 +191,7 @@ static std::vector<std::vector<int64_t>> PiperPhonemesToIdsKokoro(
current.push_back(0);

for (auto p : phonemes) {
// SHERPA_ONNX_LOGE("%d %s", static_cast<int32_t>(p), ToString(p).c_str());
if (token2id.count(p)) {
if (current.size() > max_len - 1) {
current.push_back(0);
Expand All @@ -175,6 +202,9 @@ static std::vector<std::vector<int64_t>> PiperPhonemesToIdsKokoro(
}

current.push_back(token2id.at(p));
if (p == '.') {
current.push_back(token2id.at(' '));
}
} else {
SHERPA_ONNX_LOGE("Skip unknown phonemes. Unicode codepoint: \\U+%04x.",
static_cast<uint32_t>(p));
Expand Down
Loading