Revise TTS, SpeechT5Model to end the last audio chunk at the correct …

…punctuation mark location (#513) * Revise SpeechT5Model to end the last audio chunk at the correct punctuation mark Signed-off-by: Chun Tao <[email protected]> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * still append the last chunk if it contains no punctuation Signed-off-by: Chun Tao <[email protected]> --------- Signed-off-by: Chun Tao <[email protected]> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
opea-project · Aug 19, 2024 · 20fc8ca · 20fc8ca
1 parent cd83854
commit 20fc8ca
Showing 1 changed file with 8 additions and 1 deletion.
diff --git a/comps/tts/speecht5/speecht5_model.py b/comps/tts/speecht5/speecht5_model.py
@@ -69,7 +69,14 @@ def split_long_text_into_batch(self, text, batch_length=128):
                 cur_end = idx
             idx += 1
         # deal with the last sequence
-        res.append(text[cur_start:idx])
+        if cur_start < len(text):
+            last_chunk = text[cur_start:]
+            last_punc_idx = max([last_chunk.rfind(punc) for punc in hitted_ends[:-1]])  # exclude " "
+            if last_punc_idx != -1:
+                last_chunk = last_chunk[: last_punc_idx + 1]
+                res.append(last_chunk[: last_punc_idx + 1])
+            else:
+                res.append(last_chunk)
         res = [i + "." for i in res]  # avoid unexpected end of sequence
         return res