fix: handle missing speech output in realtime demo

Bortlesboat · YaoyaoChang · commit 9fafcc1c1f03 · 2026-04-10T21:10:27.000+08:00
When outputs.speech_outputs is empty or None, the code falls through
to lines that reference audio_duration and rtf (never assigned in the
else branch), causing NameError. It also tries to index
outputs.speech_outputs[0] for save_audio, which would raise IndexError.

Early return when no audio is generated.
diff --git a/demo/realtime_model_inference_from_file.py b/demo/realtime_model_inference_from_file.py
@@ -266,7 +266,8 @@ def main():
         print(f"RTF (Real Time Factor): {rtf:.2f}x")
     else:
         print("No audio output generated")
-    
+        return
+
     # Calculate token metrics
     input_tokens = inputs['tts_text_ids'].shape[1]  # Number of input tokens
     output_tokens = outputs.sequences.shape[1]  # Total tokens (input + generated)
