diff --git a/inference_tts_scale.py b/inference_tts_scale.py
index 2ebb78c..9915b22 100644
--- a/inference_tts_scale.py
+++ b/inference_tts_scale.py
@@ -98,7 +98,9 @@ def inference_one_sample(model, model_args, phn2num, text_tokenizer, audio_token
     gen_sample = audio_tokenizer.decode(
         [(gen_frames, None)]
     )
-
+    #Empty cuda cache between runs
+    if torch.cuda.is_available():
+        torch.cuda.empty_cache()
     # return
     return concat_sample, gen_sample
 
@@ -187,4 +189,4 @@ if __name__ == "__main__":
         seg_save_fn_concat = f"{args.output_dir}/concat_{new_audio_fn[:-4]}_{i}_seed{args.seed}.wav"        
 
         torchaudio.save(seg_save_fn_gen, gen_audio, args.codec_audio_sr)
-        torchaudio.save(seg_save_fn_concat, concated_audio, args.codec_audio_sr)
\ No newline at end of file
+        torchaudio.save(seg_save_fn_concat, concated_audio, args.codec_audio_sr)