diff --git a/CHANGELOG.md b/CHANGELOG.md
index 3fb8498486..6fefa06f54 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -9,6 +9,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 
+- Added `pause_frame_processing` parameter to `CartesiaTTSService` to allow
+  disabling frame processing pausing during TTS generation.
+
 - Added `wait_for_all` argument to the base `LLMService`. When enabled, this
   ensures all function calls complete before returning results to the LLM (i.e.,
   before running a new inference with those results).
diff --git a/src/pipecat/services/cartesia/tts.py b/src/pipecat/services/cartesia/tts.py
index 9a47d92379..593defd1bf 100644
--- a/src/pipecat/services/cartesia/tts.py
+++ b/src/pipecat/services/cartesia/tts.py
@@ -234,6 +234,7 @@ def __init__(
         params: Optional[InputParams] = None,
         text_aggregator: Optional[BaseTextAggregator] = None,
         aggregate_sentences: Optional[bool] = True,
+        pause_frame_processing: Optional[bool] = True,
         **kwargs,
     ):
         """Initialize the Cartesia TTS service.
@@ -254,6 +255,7 @@ def __init__(
                     Use an LLMTextProcessor before the TTSService for custom text aggregation.
 
             aggregate_sentences: Whether to aggregate sentences within the TTSService.
+            pause_frame_processing: Whether to pause processing frames while receiving audio.
             **kwargs: Additional arguments passed to the parent service.
         """
         # Aggregating sentences still gives cleaner-sounding results and fewer
@@ -269,7 +271,7 @@ def __init__(
         super().__init__(
             aggregate_sentences=aggregate_sentences,
             push_text_frames=False,
-            pause_frame_processing=True,
+            pause_frame_processing=pause_frame_processing,
             sample_rate=sample_rate,
             text_aggregator=text_aggregator,
             **kwargs,