Add support for the Whisper model large-v3-turbo.

app.py

@@ -56,8 +56,6 @@ MAX_FILE_PREFIX_LENGTH = 17
 # Limit auto_parallel to a certain number of CPUs (specify vad_cpu_cores to get a higher number)
 MAX_AUTO_CPU_CORES = 8
 
-WHISPER_MODELS = ["tiny", "base", "small", "medium", "large", "large-v1", "large-v2", "large-v3"]
-
 class VadOptions:
     def __init__(self, vad: str = None, vadMergeWindow: float = 5, vadMaxMergeSize: float = 150, vadPadding: float = 1, vadPromptWindow: float = 1, 
                                         vadInitialPromptMode: Union[VadInitialPromptMode, str] = VadInitialPromptMode.PREPREND_FIRST_SEGMENT):

config.json5

@@ -34,6 +34,10 @@
       {
         "name": "large-v3",
         "url": "large-v3"
+      },
+      {
+        "name": "large-v3-turbo",
+        "url": "large-v3-turbo"
       }
       // Uncomment to add custom Japanese models
       //{

docs/options.md

@@ -17,6 +17,7 @@ Select the model that Whisper will use to transcribe the audio:
 | large     | 1550 M     | N/A                | large              | ~10 GB        | 1x             |
 | large-v2  | 1550 M     | N/A                | large              | ~10 GB        | 1x             |
 | large-v3  | 1550 M     | N/A                | large              | ~10 GB        | 1x             |
+| turbo     | 809 M      | N/A                | turbo              | ~6 GB         | 8x             |
 
 ## Language
 

src/translation/translationModel.py

@@ -423,6 +423,16 @@ class TranslationModel:
             else: #M2M100 & NLLB
                 output = self.transTranslator(text, max_length=max_length, batch_size=self.batchSize, no_repeat_ngram_size=self.noRepeatNgramSize, num_beams=self.numBeams)
                 result = output[0]['translation_text']
+
+            if len(result) > 2:
+                if result[len(result) - 1] == "\"" and result[0] == "\"":
+                    result = result[1:-1]
+                elif result[len(result) - 1] == "'" and result[0] == "'":
+                    result = result[1:-1]
+                elif result[len(result) - 1] == "「" and result[0] == "」":
+                    result = result[1:-1]
+                elif result[len(result) - 1] == "『" and result[0] == "』":
+                    result = result[1:-1]
         except Exception as e:
             print(traceback.format_exc())
             print("Error translation text: " + str(e))

src/whisper/fasterWhisperContainer.py

@@ -42,11 +42,13 @@ class FasterWhisperContainer(AbstractWhisperContainer):
         model_url = model_config.url
 
         if model_config.type == "whisper":
-            if model_url not in ["tiny", "base", "small", "medium", "large", "large-v1", "large-v2", "large-v3"]:
+            if model_url not in ["tiny", "base", "small", "medium", "large", "large-v1", "large-v2", "large-v3", "large-v3-turbo"]:
                 raise Exception("FasterWhisperContainer does not yet support Whisper models. Use ct2-transformers-converter to convert the model to a faster-whisper model.")
             if model_url == "large":
                 # large is an alias for large-v1
                 model_url = "large-v1"
+            elif model_url == "large-v3-turbo":
+                model_url = "deepdml/faster-whisper-large-v3-turbo-ct2"
 
         device = self.device