feat: refactor get_tts, get_speech_to_score_dict to raise immediatly errors in case of empty text/language/audio inputs

aip_trainer/lambdas/lambdaSpeechToScore.py

@@ -53,6 +53,14 @@ def get_speech_to_score_dict(real_text: str, file_bytes_or_audiotmpfile: str | d
     app_logger.info(f"real_text:{real_text} ...")
     app_logger.debug(f"file_bytes:{file_bytes_or_audiotmpfile} ...")
     app_logger.info(f"language:{language} ...")
+
+    if real_text is None or len(real_text) == 0:
+        raise ValueError(f"cannot read an empty/None text: '{real_text}'...")
+    if language is None or len(language) == 0:
+        raise NotImplementedError(f"Not tested/supported with '{language}' language...")
+    if file_bytes_or_audiotmpfile is None or len(file_bytes_or_audiotmpfile) == 0 or os.path.getsize(file_bytes_or_audiotmpfile) == 0:
+        raise ValueError(f"cannot read an empty/None file: '{file_bytes_or_audiotmpfile}'...")
+
     start0 = time.time()
 
     random_file_name = file_bytes_or_audiotmpfile

aip_trainer/lambdas/lambdaTTS.py

@@ -8,18 +8,20 @@ from aip_trainer import app_logger
 def get_tts(text: str, language: str):
     from aip_trainer.models import models
 
+    if text is None or len(text) == 0:
+        raise ValueError(f"cannot read an empty/None text: '{text}'...")
+    if language is None or len(language) == 0:
+        raise NotImplementedError(f"Not tested/supported with '{language}' language...")
+
     tmp_dir = Path(tempfile.gettempdir())
-    if language == "en" or language == "de":
-        try:
-            model, _, speaker, sample_rate = models.silero_tts(
-                language, output_folder=tmp_dir
-            )
-        except ValueError:
-            model, _, sample_rate, _, _, speaker = models.silero_tts(
-                language, output_folder=tmp_dir
-            )
-    else:
-        raise NotImplementedError(f"Not yet tested with {language} error...")
+    try:
+        model, _, speaker, sample_rate = models.silero_tts(
+            language, output_folder=tmp_dir
+        )
+    except ValueError:
+        model, _, sample_rate, _, _, speaker = models.silero_tts(
+            language, output_folder=tmp_dir
+        )
     app_logger.info(f"model speaker #0: {speaker} ...")
 
     with tempfile.NamedTemporaryFile(prefix="audio_", suffix=".wav", delete=False) as tmp_audio_file:
@@ -27,26 +29,3 @@ def get_tts(text: str, language: str):
         audio_paths = model.save_wav(text=text, speaker=speaker, sample_rate=sample_rate, audio_path=str(tmp_audio_file.name))
         app_logger.info(f"audio_paths output: {audio_paths} ...")
         return audio_paths
-
-
-"""
-Help on method save_wav:
-
-save_wav(text=None, ssml_text=None, speaker: str = 'xenia', audio_path: str = '', sample_rate: int = 48000, put_accent=True, put_yo=True) method of <torch_package_0>.multi_acc_v3_package.TTSModelMultiAcc_v3 instance
-
-
-    tmp_dir = tempfile.gettempdir()
-    if language == "de":
-        model, decoder, _ = silero_stt(
-            language="de", version="v4", jit_model="jit_large", output_folder=tmp_dir
-        )
-    elif language == "en":
-        model, decoder, _ = silero_stt(language="en", output_folder=tmp_dir)
-    else:
-        raise NotImplementedError(
-            "currenty works only for 'de' and 'en' languages, not for '{}'.".format(
-                language
-            )
-        )
-
-"""