fixes

app.py

@@ -88,9 +88,9 @@ def text_to_speech(text):
     return text, (16000, audio.squeeze())
 
 
-def transcribe(image, audio):
+def transcribe_pipeline(image, audio):
     if not image or not audio:
-        return
+        return None, None
 
     sr, y = audio
 
@@ -102,11 +102,18 @@ def transcribe(image, audio):
 
     transcription_text = transcriber({"sampling_rate": sr, "raw": y})["text"]
 
-    return generate_answer(image, transcription_text)
+    return text_to_speech(generate_answer(image, transcription_text))
+
+
+def text_pipeline(image, question):
+    if not image or not question:
+        return None, None
+
+    return text_to_speech(generate_answer(image, question))
 
 
 qa_interface = gr.Interface(
-    fn=generate_answer,
+    fn=text_pipeline,
     inputs=[
         gr.Image(type="pil"),
         gr.Textbox(label="Вопрос (на русском)", placeholder="Ваш вопрос"),
@@ -120,7 +127,7 @@ qa_interface = gr.Interface(
 )
 
 speech_interface = gr.Interface(
-    fn=transcribe,
+    fn=transcribe_pipeline,
     inputs=[
         gr.Image(type="pil"),
         gr.Audio(sources="microphone", label="Голосовой ввод"),