Update app.py

app.py

@@ -9,10 +9,8 @@ import wave
 from balacoon_tts import TTS
 from threading import Lock
 from huggingface_hub import hf_hub_download, list_repo_files
-from pydub import AudioSegment
 import io
 import tempfile
-import pydub
 
 #tts cpu model
 tts_model_str = "en_us_hifi_jets_cpu.addon"
@@ -47,12 +45,19 @@ def text_to_speech(text):
     return output_file
 
 def combine_audio_files(audio_files):
-    combined = AudioSegment.empty()
-    for audio_file in audio_files:
-        segment = AudioSegment.from_wav(audio_file)
-        combined += segment
-        os.remove(audio_file)  # Remove temporary files
-    return combined
+    data= []
+    outfile = "sounds.wav"
+    for infile in audio_files:
+        w = wave.open(infile, 'rb')
+        data.append([w.getparams(), w.readframes(w.getnframes())] )
+        w.close()
+        os.remove(infile)  # Remove temporary files
+    output = wave.open(outfile, 'wb')
+    output.setparams(data[0][0])
+    for i in range(len(data)):
+        output.writeframes(data[i][1])
+    output.close()
+    return outfile
     
 #client
 client = Groq(
@@ -127,7 +132,7 @@ async def greet(product,description):
     if response.choices[0].message.content != "not moderated":
         a_list = ["Sorry, I can't proceed for generating marketing email. Your content needs to be moderated first. Thank you!"]
         processed_audio = combine_audio_files([text_to_speech(a_list[0])])
-        yield (processed_audio.frame_rate,processed_audio), a_list[0]
+        yield processed_audio, a_list[0]
     else:
         output = llm.create_chat_completion(
         messages=[
@@ -149,7 +154,7 @@ async def greet(product,description):
                 audio_list = audio_list + [text_to_speech(delta.get('content', ''))]
                 processed_audio = combine_audio_files(audio_list)
                 partial_message = partial_message + delta.get('content', '')
-                yield (processed_audio.frame_rate,processed_audio), partial_message
+                yield processed_audio, partial_message
 
 audio = gr.Audio()
 demo = gr.Interface(fn=greet, inputs=["text","text"], concurrency_limit=10, outputs=[audio,"text"])