fix

ultravox_processing.py

@@ -180,7 +180,7 @@ class UltravoxProcessor(transformers.ProcessorMixin):
 
                     start_idx = len(
                         self.tokenizer.encode(
-                            t[: text.t(self.audio_placeholder)],
+                            t[: t.index(self.audio_placeholder)],
                             add_special_tokens=False,
                         )
                     )