Add language

app.py

@@ -8,6 +8,7 @@ with gr.Blocks() as demo:
     model = gr.Dropdown(value='large-v3', choices=["tiny", "base", "small", "medium", "large", "large-v2", "large-v3"])
     run_button = gr.Button(value='Run')
     prompt = gr.Textbox(label='prompt')
+    language = gr.Textbox(label='language')
     set_button = gr.Button(value='Set Prompt')
     text_only = gr.Textbox(label='output')
     text_with_timestamps = gr.Textbox(label='timestamps')
@@ -20,7 +21,7 @@ with gr.Blocks() as demo:
 
     set_button.click(
         fn=fn.set_prompt,
-        inputs=[prompt],
+        inputs=[prompt, language],
         outputs=[],
     )
 

fn.py

@@ -3,6 +3,7 @@ from faster_whisper import WhisperModel
 model = None
 model_size = None
 initial_prompt = None
+language = 'ja'
 
 def load_model(_model_size):
     global model_size, model
@@ -15,9 +16,11 @@ def load_model(_model_size):
     except:
         model = WhisperModel(model_size, device="cpu", compute_type="int8")
 
-def set_prompt(prompt):
-    global initial_prompt
+def set_prompt(prompt, _language = None):
+    global initial_prompt, language
     initial_prompt = prompt
+    if _language:
+        language = _language
 
 def speech_to_text(audio_file, _model_size = None):
     global model_size, model
@@ -27,7 +30,7 @@ def speech_to_text(audio_file, _model_size = None):
     segments, info = model.transcribe(
         audio_file,
         initial_prompt=initial_prompt,
-        language='ja',
+        language=language,
         beam_size=5,
         vad_filter=True,
         without_timestamps=False,

main.py

@@ -42,9 +42,9 @@ async def transcribe_audio(file: UploadFile = Form(...)):
         return {"error": str(e)}
 
 @app.post("/set_prompt")
-async def set_prompt(prompt: str):
+async def set_prompt(prompt: str, language: str = None):
     try:
-        fn.set_prompt(prompt)
+        fn.set_prompt(prompt, language)
 
         return {"status": 0}
     except Exception as e: