dv-pipecat-ai 0.0.85.dev831__py3-none-any.whl → 0.0.85.dev833__py3-none-any.whl

{dv_pipecat_ai-0.0.85.dev831.dist-info → dv_pipecat_ai-0.0.85.dev833.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dv-pipecat-ai
-Version: 0.0.85.dev831
+Version: 0.0.85.dev833
 Summary: An open source framework for voice (and multimodal) assistants
 License-Expression: BSD-2-Clause
 Project-URL: Source, https://github.com/pipecat-ai/pipecat

{dv_pipecat_ai-0.0.85.dev831.dist-info → dv_pipecat_ai-0.0.85.dev833.dist-info}/RECORD

@@ -1,4 +1,4 @@
-dv_pipecat_ai-0.0.85.dev831.dist-info/licenses/LICENSE,sha256=DWY2QGf2eMCFhuu2ChairtT6CB7BEFffNVhXWc4Od08,1301
+dv_pipecat_ai-0.0.85.dev833.dist-info/licenses/LICENSE,sha256=DWY2QGf2eMCFhuu2ChairtT6CB7BEFffNVhXWc4Od08,1301
 pipecat/__init__.py,sha256=j0Xm6adxHhd7D06dIyyPV_GlBYLlBnTAERVvD_jAARQ,861
 pipecat/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 pipecat/adapters/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -205,7 +205,7 @@ pipecat/services/azure/realtime/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm
 pipecat/services/azure/realtime/llm.py,sha256=MnDiw-YJP3kll1gbkta4z4vsWfWZ5oBprZCinMP9O0M,2385
 pipecat/services/cartesia/__init__.py,sha256=vzh0jBnfPwWdxFfV-tu0x1HFoOTgr9s91GYmD-CJUtY,284
 pipecat/services/cartesia/stt.py,sha256=00k9gQYo_xPKb-RRJ-RNV4LPFw-7xXiFU7ACFLYttWY,12388
-pipecat/services/cartesia/tts.py,sha256=Fh6hm5AUj2rNX8J4UOjHA7uAPIGcie1Dyxv5WBvV1OY,26279
+pipecat/services/cartesia/tts.py,sha256=I_OZCINywkDXmYzFL35MjSN8cAuNEaJs7nj0YB_obtc,27008
 pipecat/services/cerebras/__init__.py,sha256=5zBmqq9Zfcl-HC7ylekVS5qrRedbl1mAeEwUT-T-c_o,259
 pipecat/services/cerebras/llm.py,sha256=-yzSe_6YDGigwzES-LZS4vNXMPugmvsIYEpTySyr5nA,3047
 pipecat/services/deepgram/__init__.py,sha256=IjRtMI7WytRDdmYVpk2qDWClXUiNgdl7ZkvEAWg1eYE,304
@@ -353,7 +353,7 @@ pipecat/transcriptions/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3
 pipecat/transcriptions/language.py,sha256=-mWI1MiZbasuoqZTOBH69dAmoM7-UJzWq9rSCcrnmh4,8228
 pipecat/transports/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 pipecat/transports/base_input.py,sha256=WGtoXXlF3GIjYgjtYnAgi8nZozd5abNlGNjwRnz8FRs,20138
-pipecat/transports/base_output.py,sha256=mNlIOo7tETlbYPbDyOtA2H-TkBGFKmjuCMDzQUtiwmk,35423
+pipecat/transports/base_output.py,sha256=7WoXtAQAi-3OC9PC_zk61lCWlBTk5-NuTLUbsQUAI_U,36723
 pipecat/transports/base_transport.py,sha256=JlNiH0DysTfr6azwHauJqY_Z9HJC702O29Q0qrsLrg4,7530
 pipecat/transports/daily/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 pipecat/transports/daily/transport.py,sha256=VanO33ff9g6px-vwGgT6M7cMVg786pOGfMU7Okm7a78,91917
@@ -415,7 +415,7 @@ pipecat/utils/tracing/service_decorators.py,sha256=fwzxFpi8DJl6BJbK74G0UEB4ccMJg
 pipecat/utils/tracing/setup.py,sha256=7TEgPNpq6M8lww8OQvf0P9FzYc5A30xICGklVA-fua0,2892
 pipecat/utils/tracing/turn_context_provider.py,sha256=ikon3plFOx0XbMrH6DdeHttNpb-U0gzMZIm3bWLc9eI,2485
 pipecat/utils/tracing/turn_trace_observer.py,sha256=dma16SBJpYSOE58YDWy89QzHyQFc_9gQZszKeWixuwc,9725
-dv_pipecat_ai-0.0.85.dev831.dist-info/METADATA,sha256=5ahEs864DAPIEEiiv7-7Oa-vRhRN1Ede341NuqED3Sw,32924
-dv_pipecat_ai-0.0.85.dev831.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-dv_pipecat_ai-0.0.85.dev831.dist-info/top_level.txt,sha256=kQzG20CxGf-nSsHmtXHx3hY2-8zHA3jYg8jk0TajqXc,8
-dv_pipecat_ai-0.0.85.dev831.dist-info/RECORD,,
+dv_pipecat_ai-0.0.85.dev833.dist-info/METADATA,sha256=8G_xwlvsUOMNtLW1-haMrK98GNcgh0tMxIuPx3s7aQk,32924
+dv_pipecat_ai-0.0.85.dev833.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+dv_pipecat_ai-0.0.85.dev833.dist-info/top_level.txt,sha256=kQzG20CxGf-nSsHmtXHx3hY2-8zHA3jYg8jk0TajqXc,8
+dv_pipecat_ai-0.0.85.dev833.dist-info/RECORD,,

pipecat/services/cartesia/tts.py

@@ -93,6 +93,33 @@ def language_to_cartesia_language(language: Language) -> Optional[str]:
         Language.SV: "sv",
         Language.TR: "tr",
         Language.ZH: "zh",
+        Language.TL: "tl",
+        Language.BG: "bg",
+        Language.RO: "ro",
+        Language.AR: "ar",
+        Language.CS: "cs",
+        Language.EL: "el",
+        Language.FI: "fi",
+        Language.HR: "hr",
+        Language.MS: "ms",
+        Language.SK: "sk",
+        Language.DA: "da",
+        Language.TA: "ta",
+        Language.UK: "uk",
+        Language.HU: "hu",
+        Language.NO: "no",
+        Language.VI: "vi",
+        Language.BN: "bn",
+        Language.TH: "th",
+        Language.HE: "he",
+        Language.KA: "ka",
+        Language.ID: "id",
+        Language.TE: "te",
+        Language.GU: "gu",
+        Language.KN: "kn",
+        Language.ML: "ml",
+        Language.MR: "mr",
+        Language.PA: "pa",
     }
 
     result = BASE_LANGUAGES.get(language)

pipecat/transports/base_output.py

@@ -50,6 +50,11 @@ from pipecat.utils.time import nanoseconds_to_seconds
 
 # TODO: When we use GeminiMultimodalLiveLLMService, we need to change this to 0.35 but that creates issue for faster TTS.
 BOT_VAD_STOP_SECS = 0.30
+# For the very first bot utterance (e.g., intro), we can safely
+# detect end-of-speech sooner to improve responsiveness for the
+# user’s first short reply. Keep conservative to avoid mid-utterance
+# false stops when TTS streams quickly.
+FIRST_BOT_VAD_STOP_SECS = 0.08
 
 
 class BaseOutputTransport(FrameProcessor):
@@ -406,6 +411,9 @@ class BaseOutputTransport(FrameProcessor):
             self._bot_speaking_frame_period = 0.2
             # Last time the bot actually spoke.
             self._bot_speech_last_time = 0
+            # Before the first stop event, we use a shorter silence
+            # threshold to make the first turn more responsive.
+            self._first_stop_pending = True
 
             self._audio_task: Optional[asyncio.Task] = None
             self._video_task: Optional[asyncio.Task] = None
@@ -631,6 +639,10 @@ class BaseOutputTransport(FrameProcessor):
 
             self._bot_speaking = False
 
+            # Mark that the first stop has been completed so subsequent
+            # stops use the regular (longer) VAD stop threshold.
+            self._first_stop_pending = False
+
             # Clean audio buffer (there could be tiny left overs if not multiple
             # to our output chunk size).
             self._audio_buffer = bytearray()
@@ -690,9 +702,14 @@ class BaseOutputTransport(FrameProcessor):
             async def without_mixer(vad_stop_secs: float) -> AsyncGenerator[Frame, None]:
                 while True:
                     try:
-                        frame = await asyncio.wait_for(
-                            self._audio_queue.get(), timeout=vad_stop_secs
+                        # Use a shorter timeout only for the first bot stop to
+                        # accelerate the initial turn handoff right after the intro.
+                        timeout = (
+                            FIRST_BOT_VAD_STOP_SECS
+                            if getattr(self, "_first_stop_pending", True)
+                            else BOT_VAD_STOP_SECS
                         )
+                        frame = await asyncio.wait_for(self._audio_queue.get(), timeout=timeout)
                         yield frame
                         self._audio_queue.task_done()
                     except asyncio.TimeoutError:
@@ -713,7 +730,13 @@ class BaseOutputTransport(FrameProcessor):
                     except asyncio.QueueEmpty:
                         # Notify the bot stopped speaking upstream if necessary.
                         diff_time = time.time() - last_frame_time
-                        if diff_time > vad_stop_secs:
+                        # Use a shorter threshold for the first stop only.
+                        current_stop_secs = (
+                            FIRST_BOT_VAD_STOP_SECS
+                            if getattr(self, "_first_stop_pending", True)
+                            else BOT_VAD_STOP_SECS
+                        )
+                        if diff_time > current_stop_secs:
                             await self._bot_stopped_speaking()
                         # Generate an audio frame with only the mixer's part.
                         frame = OutputAudioRawFrame(