Add code

app.py

@@ -10,7 +10,8 @@ import librosa
 
 pipe = transformers.pipeline(model='fixie-ai/ultravox-v0_4_1-llama-3_1-8b', trust_remote_code=True,
                              device=torch.device('cuda'))
-
+whisper = transformers.pipeline(model="openai/whisper-large-v3-turbo",
+                                device=torch.device('cuda'))
 
 account_sid = os.environ.get("TWILIO_ACCOUNT_SID")
 auth_token = os.environ.get("TWILIO_AUTH_TOKEN")
@@ -35,16 +36,14 @@ def transcribe(audio: tuple[int, np.ndarray], conversation: list[dict], gradio_c
 
     audio_sr = librosa.resample(audio[1].astype(np.float32) / 32768.0,
                                 orig_sr=original_sr, target_sr=target_sr)
-    inputs = pipe.preprocess({"audio": audio_sr, "turns": conversation, "sampling_rate": target_sr})
-    print("preprocess", inputs)
    
     output = pipe({"audio": audio_sr, "turns": conversation, "sampling_rate": target_sr},
                   max_new_tokens=512)
-    print("output", output)
+    transcription = whisper({"array": audio_sr, "sampling_rate": target_sr})
 
-    conversation.append({"role": "user", "content": "hi"})
+    conversation.append({"role": "user", "content": transcription})
     conversation.append({"role": "assistant", "content": output})
-    gradio_convo.append({"role": "user", "content": "hi"})
+    gradio_convo.append({"role": "user", "content": transcription})
     gradio_convo.append({"role": "assistant", "content": output})
 
     yield AdditionalOutputs(conversation, gradio_convo)