cartesia 1.1.0.dev0__py3-none-any.whl → 1.3.0__py3-none-any.whl

cartesia/_async_websocket.py

@@ -6,7 +6,7 @@ from typing import Any, AsyncGenerator, Callable, Dict, List, Optional, Union
 
 import aiohttp
 
-from cartesia._constants import DEFAULT_MODEL_ID, DEFAULT_VOICE_EMBEDDING
+from cartesia._constants import DEFAULT_MODEL_ID, DEFAULT_OUTPUT_FORMAT, DEFAULT_VOICE_EMBEDDING
 from cartesia._types import OutputFormat, VoiceControls
 from cartesia._websocket import _WebSocket
 from cartesia.tts import TTS
@@ -45,6 +45,7 @@ class _AsyncTTSContext:
         voice_embedding: Optional[List[float]] = None,
         context_id: Optional[str] = None,
         continue_: bool = False,
+        flush: bool = False,
         duration: Optional[int] = None,
         language: Optional[str] = None,
         add_timestamps: bool = False,
@@ -60,6 +61,7 @@ class _AsyncTTSContext:
             voice_embedding: The embedding of the voice to use for generating audio.
             context_id: The context ID to use for the request. If not specified, a random context ID will be generated.
             continue_: Whether to continue the audio generation from the previous transcript or not.
+            flush: Whether to trigger a manual flush for the current context's generation.
             duration: The duration of the audio in seconds.
             language: The language code for the audio request. This can only be used with `model_id = sonic-multilingual`.
             add_timestamps: Whether to return word-level timestamps.
@@ -71,7 +73,7 @@ class _AsyncTTSContext:
         """
         if context_id is not None and context_id != self._context_id:
             raise ValueError("Context ID does not match the context ID of the current context.")
-        if continue_ and transcript == "":
+        if continue_ and transcript == "" and not flush:
             raise ValueError("Transcript cannot be empty when continue_ is True.")
 
         await self._websocket.connect()
@@ -87,6 +89,7 @@ class _AsyncTTSContext:
             context_id=self._context_id,
             add_timestamps=add_timestamps,
             continue_=continue_,
+            flush=flush,
             _experimental_voice_controls=_experimental_voice_controls,
         )
 
@@ -100,12 +103,49 @@ class _AsyncTTSContext:
         await self.send(
             model_id=DEFAULT_MODEL_ID,
             transcript="",
-            output_format=TTS.get_output_format("raw_pcm_f32le_44100"),
+            output_format=TTS.get_output_format(DEFAULT_OUTPUT_FORMAT),
             voice_embedding=DEFAULT_VOICE_EMBEDDING,  # Default voice embedding since it's a required input for now.
             context_id=self._context_id,
             continue_=False,
         )
 
+    async def flush(self) -> Callable[[], AsyncGenerator[Dict[str, Any], None]]:
+        """Trigger a manual flush for the current context's generation. This method returns a generator that yields the audio prior to the flush."""
+        await self.send(
+            model_id=DEFAULT_MODEL_ID,
+            transcript="",
+            output_format=TTS.get_output_format(DEFAULT_OUTPUT_FORMAT),
+            voice_embedding=DEFAULT_VOICE_EMBEDDING,  # Default voice embedding since it's a required input for now.
+            context_id=self._context_id,
+            continue_=True,
+            flush=True,
+        )
+
+        # Save the old flush ID
+        flush_id = len(self._websocket._context_queues[self._context_id]) - 1
+
+        # Create a new Async Queue to store the responses for the new flush ID
+        self._websocket._context_queues[self._context_id].append(asyncio.Queue())
+
+        # Return the generator for the old flush ID
+        async def generator():
+            try:
+                while True:
+                    response = await self._websocket._get_message(
+                        self._context_id, timeout=self.timeout, flush_id=flush_id
+                    )
+                    if "error" in response:
+                        raise RuntimeError(f"Error generating audio:\n{response['error']}")
+                    if response.get("flush_done") or response["done"]:
+                        break
+                    yield self._websocket._convert_response(response, include_context_id=True)
+            except Exception as e:
+                if isinstance(e, asyncio.TimeoutError):
+                    raise RuntimeError("Timeout while waiting for audio chunk")
+                raise RuntimeError(f"Failed to generate audio:\n{e}")
+
+        return generator
+
     async def receive(self) -> AsyncGenerator[Dict[str, Any], None]:
         """Receive the audio chunks from the WebSocket. This method is a generator that yields audio chunks.
 
@@ -175,7 +215,7 @@ class _AsyncWebSocket(_WebSocket):
         self.timeout = timeout
         self._get_session = get_session
         self.websocket = None
-        self._context_queues: Dict[str, asyncio.Queue] = {}
+        self._context_queues: Dict[str, List[asyncio.Queue]] = {}
         self._processing_task: asyncio.Task = None
 
     def __del__(self):
@@ -213,7 +253,7 @@ class _AsyncWebSocket(_WebSocket):
             except asyncio.CancelledError:
                 pass
             except TypeError as e:
-                # Ignore the error if the task is already cancelled
+                # Ignore the error if the task is already canceled.
                 # For some reason we are getting None responses
                 # TODO: This needs to be fixed - we need to think about why we are getting None responses.
                 if "Received message 256:None" not in str(e):
@@ -284,16 +324,23 @@ class _AsyncWebSocket(_WebSocket):
                 response = await self.websocket.receive_json()
                 if response["context_id"]:
                     context_id = response["context_id"]
+                flush_id = response.get("flush_id", -1)
                 if context_id in self._context_queues:
-                    await self._context_queues[context_id].put(response)
+                    await self._context_queues[context_id][flush_id].put(response)
         except Exception as e:
             self._error = e
             raise e
 
-    async def _get_message(self, context_id: str, timeout: float) -> Dict[str, Any]:
+    async def _get_message(
+        self, context_id: str, timeout: float, flush_id: Optional[int] = -1
+    ) -> Dict[str, Any]:
         if context_id not in self._context_queues:
             raise ValueError(f"Context ID {context_id} not found.")
-        return await asyncio.wait_for(self._context_queues[context_id].get(), timeout=timeout)
+        if len(self._context_queues[context_id]) <= flush_id:
+            raise ValueError(f"Flush ID {flush_id} not found for context ID {context_id}.")
+        return await asyncio.wait_for(
+            self._context_queues[context_id][flush_id].get(), timeout=timeout
+        )
 
     def _remove_context(self, context_id: str):
         if context_id in self._context_queues:
@@ -309,5 +356,5 @@ class _AsyncWebSocket(_WebSocket):
         if context_id is None:
             context_id = str(uuid.uuid4())
         if context_id not in self._context_queues:
-            self._context_queues[context_id] = asyncio.Queue()
+            self._context_queues[context_id] = [asyncio.Queue()]
         return _AsyncTTSContext(context_id, self, self.timeout)

cartesia/_constants.py

@@ -2,6 +2,7 @@ DEFAULT_MODEL_ID = "sonic-english"  # latest default model
 MULTILINGUAL_MODEL_ID = "sonic-multilingual"  # latest multilingual model
 DEFAULT_BASE_URL = "api.cartesia.ai"
 DEFAULT_CARTESIA_VERSION = "2024-06-10"  # latest version
+DEFAULT_OUTPUT_FORMAT = "raw_pcm_f32le_44100"
 DEFAULT_TIMEOUT = 30  # seconds
 DEFAULT_NUM_CONNECTIONS = 10  # connections per client
 DEFAULT_VOICE_EMBEDDING = [1.0] * 192

cartesia/_types.py

@@ -27,39 +27,6 @@ class OutputFormatMapping:
             raise ValueError(f"Unsupported format: {format_name}")
 
 
-class DeprecatedOutputFormatMapping:
-    """Deprecated formats as of v1.0.1. These will be removed in v1.2.0. Use :class:`OutputFormatMapping` instead."""
-
-    _format_mapping = {
-        "fp32": {"container": "raw", "encoding": "pcm_f32le", "sample_rate": 44100},
-        "pcm": {"container": "raw", "encoding": "pcm_s16le", "sample_rate": 44100},
-        "fp32_8000": {"container": "raw", "encoding": "pcm_f32le", "sample_rate": 8000},
-        "fp32_16000": {"container": "raw", "encoding": "pcm_f32le", "sample_rate": 16000},
-        "fp32_22050": {"container": "raw", "encoding": "pcm_f32le", "sample_rate": 22050},
-        "fp32_24000": {"container": "raw", "encoding": "pcm_f32le", "sample_rate": 24000},
-        "fp32_44100": {"container": "raw", "encoding": "pcm_f32le", "sample_rate": 44100},
-        "pcm_8000": {"container": "raw", "encoding": "pcm_s16le", "sample_rate": 8000},
-        "pcm_16000": {"container": "raw", "encoding": "pcm_s16le", "sample_rate": 16000},
-        "pcm_22050": {"container": "raw", "encoding": "pcm_s16le", "sample_rate": 22050},
-        "pcm_24000": {"container": "raw", "encoding": "pcm_s16le", "sample_rate": 24000},
-        "pcm_44100": {"container": "raw", "encoding": "pcm_s16le", "sample_rate": 44100},
-        "mulaw_8000": {"container": "raw", "encoding": "pcm_mulaw", "sample_rate": 8000},
-        "alaw_8000": {"container": "raw", "encoding": "pcm_alaw", "sample_rate": 8000},
-    }
-
-    @classmethod
-    @deprecated(
-        vdeprecated="1.0.1",
-        vremove="1.2.0",
-        reason="Old output format names are being deprecated in favor of names aligned with the Cartesia API. Use names from `OutputFormatMapping` instead.",
-    )
-    def get_format_deprecated(cls, format_name):
-        if format_name in cls._format_mapping:
-            return cls._format_mapping[format_name]
-        else:
-            raise ValueError(f"Unsupported format: {format_name}")
-
-
 class VoiceMetadata(TypedDict):
     id: str
     name: str

cartesia/_websocket.py

@@ -239,7 +239,7 @@ class _WebSocket:
             self._contexts.clear()
 
     def _convert_response(
-        self, response: Dict[str, any], include_context_id: bool
+        self, response: Dict[str, any], include_context_id: bool, include_flush_id: bool = False
     ) -> Dict[str, Any]:
         out = {}
         if response["type"] == EventType.AUDIO:
@@ -250,6 +250,9 @@ class _WebSocket:
         if include_context_id:
             out["context_id"] = response["context_id"]
 
+        if include_flush_id and "flush_id" in response:
+            out["flush_id"] = response["flush_id"]
+
         return out
 
     def send(

cartesia/tts.py

@@ -4,7 +4,6 @@ import httpx
 
 from cartesia._sse import _SSE
 from cartesia._types import (
-    DeprecatedOutputFormatMapping,
     OutputFormat,
     OutputFormatMapping,
     VoiceControls,
@@ -86,10 +85,6 @@ class TTS(Resource):
         """
         if output_format_name in OutputFormatMapping._format_mapping:
             output_format_obj = OutputFormatMapping.get_format(output_format_name)
-        elif output_format_name in DeprecatedOutputFormatMapping._format_mapping:
-            output_format_obj = DeprecatedOutputFormatMapping.get_format_deprecated(
-                output_format_name
-            )
         else:
             raise ValueError(f"Unsupported format: {output_format_name}")
 
@@ -114,10 +109,6 @@ class TTS(Resource):
         """
         if output_format_name in OutputFormatMapping._format_mapping:
             output_format_obj = OutputFormatMapping.get_format(output_format_name)
-        elif output_format_name in DeprecatedOutputFormatMapping._format_mapping:
-            output_format_obj = DeprecatedOutputFormatMapping.get_format_deprecated(
-                output_format_name
-            )
         else:
             raise ValueError(f"Unsupported format: {output_format_name}")
 

cartesia/utils/tts.py

@@ -37,6 +37,7 @@ def _construct_tts_request(
     add_timestamps: bool = False,
     context_id: Optional[str] = None,
     continue_: bool = False,
+    flush: bool = False,
     _experimental_voice_controls: Optional[VoiceControls] = None,
 ):
     tts_request = {
@@ -71,4 +72,7 @@ def _construct_tts_request(
     if continue_:
         tts_request["continue"] = continue_
 
+    if flush:
+        tts_request["flush"] = flush
+
     return tts_request

cartesia/version.py

@@ -1 +1 @@
-__version__ = "1.1.0-dev0"
+__version__ = "1.3.0"

cartesia/voices.py

@@ -58,29 +58,65 @@ class Voices(Resource):
 
         return response.json()
 
-    def clone(self, filepath: Optional[str] = None, enhance: str = True) -> List[float]:
+    def clone(
+        self,
+        filepath: Optional[str] = None,
+        enhance: str = True,
+        mode: str = "clip",
+        language: str = "en",
+        name: Optional[str] = None,
+        description: Optional[str] = None,
+        transcript: Optional[str] = None,
+    ) -> Union[List[float], VoiceMetadata]:
         """Clone a voice from a clip.
 
         Args:
             filepath: The path to the clip file.
             enhance: Whether to enhance the clip before cloning the voice (highly recommended). Defaults to True.
+            mode: The mode to use for cloning. Either "similarity" or "stability".
+            language: The language code of the language spoken in the clip. Defaults to "en".
+            name: The name of the cloned voice.
+            description: The description of the cloned voice.
+            transcript: The transcript of the clip. Only used if mode is "similarity".
 
         Returns:
             The embedding of the cloned voice as a list of floats.
         """
         if not filepath:
             raise ValueError("Filepath must be specified.")
-        url = f"{self._http_url()}/voices/clone/clip"
+        headers = self.headers.copy()
+        headers.pop("Content-Type", None)
+
         with open(filepath, "rb") as file:
             files = {"clip": file}
-            files["enhance"] = str(enhance).lower()
-            headers = self.headers.copy()
-            headers.pop("Content-Type", None)
-            response = httpx.post(url, headers=headers, files=files, timeout=self.timeout)
-            if not response.is_success:
-                raise ValueError(f"Failed to clone voice from clip. Error: {response.text}")
-
-        return response.json()["embedding"]
+            data = {
+                "enhance": str(enhance).lower(),
+                "mode": mode,
+            }
+            if mode == "clip":
+                url = f"{self._http_url()}/voices/clone/clip"
+                response = httpx.post(
+                    url, headers=headers, files=files, data=data, timeout=self.timeout
+                )
+                if not response.is_success:
+                    raise ValueError(f"Failed to clone voice from clip. Error: {response.text}")
+                return response.json()["embedding"]
+            else:
+                data["name"] = name
+                data["description"] = description
+                data["language"] = language
+                if mode == "similarity" and transcript:
+                    data["transcript"] = transcript
+                url = f"{self._http_url()}/voices/clone"
+                response = httpx.post(
+                    url, headers=headers, files=files, data=data, timeout=self.timeout
+                )
+                if not response.is_success:
+                    raise ValueError(
+                        f"Failed to clone voice. Status Code: {response.status_code}\n"
+                        f"Error: {response.text}"
+                    )
+                return response.json()
 
     def create(
         self,
@@ -88,6 +124,7 @@ class Voices(Resource):
         description: str,
         embedding: List[float],
         base_voice_id: Optional[str] = None,
+        language: str = "en",
     ) -> VoiceMetadata:
         """Create a new voice.
 
@@ -108,6 +145,7 @@ class Voices(Resource):
                 "description": description,
                 "embedding": embedding,
                 "base_voice_id": base_voice_id,
+                "language": language,
             },
             timeout=self.timeout,
         )

{cartesia-1.1.0.dev0.dist-info → cartesia-1.3.0.dist-info}/METADATA

@@ -1,15 +1,14 @@
-Metadata-Version: 2.1
+Metadata-Version: 2.3
 Name: cartesia
-Version: 1.1.0.dev0
+Version: 1.3.0
 Summary: The official Python library for the Cartesia API.
 Requires-Python: >=3.9
+Requires-Dist: aiohttp>=3.10.10
+Requires-Dist: httpx>=0.27.2
+Requires-Dist: iterators>=0.2.0
+Requires-Dist: requests>=2.32.3
+Requires-Dist: websockets>=10.4
 Description-Content-Type: text/markdown
-License-File: LICENSE.md
-Requires-Dist: aiohttp >=3.10.10
-Requires-Dist: httpx >=0.27.2
-Requires-Dist: iterators >=0.2.0
-Requires-Dist: requests >=2.32.3
-Requires-Dist: websockets >=13.1
 
 # Cartesia Python API Library
 
@@ -642,8 +641,6 @@ display(audio)
 
 You can use the `client.tts.get_output_format` method to convert string-based output format names into the `output_format` dictionary which is expected by the `output_format` parameter. You can see the `OutputFormatMapping` class in `cartesia._types` for the currently supported output format names. You can also view the currently supported `output_format`s in our [API Reference](https://docs.cartesia.ai/reference/api-reference/rest/stream-speech-server-sent-events).
 
-The previously used `output_format` strings are now deprecated and will be removed in v1.2.0. These are listed in the `DeprecatedOutputFormatMapping` class in `cartesia._types`.
-
 ```python
 # Get the output format dictionary from string name
 output_format = client.tts.get_output_format("raw_pcm_f32le_44100")

cartesia-1.3.0.dist-info/RECORD

@@ -0,0 +1,23 @@
+cartesia/__init__.py,sha256=rS7jIg4iqT0VgnwjzYK25JXxnF5hjZGE_-PGynAqHFo,126
+cartesia/_async_sse.py,sha256=76oIvstzVcWZCbcD8Ps419k1FEHF6lOB5qoHwawvj9k,3327
+cartesia/_async_websocket.py,sha256=y9YL9fU8eLENZZECJUwRBVTfEx4ZMl96Y5zHaRY2BiI,14787
+cartesia/_constants.py,sha256=khGNVpiQVDmv1oZU7pKTd9C1AHjiaM8zQ2He9d5zI_c,435
+cartesia/_logger.py,sha256=vU7QiGSy_AJuJFmClUocqIJ-Ltku_8C24ZU8L6fLJR0,53
+cartesia/_sse.py,sha256=CugabGUAUM-N2BruxNFxDB20HyxDlRdbN-J_yAzvBMY,5667
+cartesia/_types.py,sha256=gixQbKbX-H8xbD7jxHmc02KXLyjEaup19lh_57_YBl8,2570
+cartesia/_websocket.py,sha256=nRCq9xB0T9yYHoLqtn0GsJmcap-OAlJdSIrzTl40qMI,14875
+cartesia/async_client.py,sha256=y_K_Yuv0weA4k9ZYD0M9bNM3x3frsq07tqkg7R9h0-o,2714
+cartesia/async_tts.py,sha256=IbWVRKklNClXASR6ylHaukcMRR304LUguqc4yMopbDU,2076
+cartesia/client.py,sha256=OS1ORUSlR8Jg-em1imeTAFfwkC85AQFnw8PYtTdUuC8,2364
+cartesia/resource.py,sha256=wpnB3IPcTdxYSp0vxSkpntp4NSvqvnwUWF-0ZpgWV9o,1585
+cartesia/tts.py,sha256=kWvqce9K3gZ4QrWD-ciYdK29n49SNkxhd2A7ueTOwMY,4878
+cartesia/version.py,sha256=F5mW07pSyGrqDNY2Ehr-UpDzpBtN-FsYU0QGZWf6PJE,22
+cartesia/voices.py,sha256=bDYbs0KoikAROJlmbnLdo4TrW0YwzjMvp70uKG6Alp0,7180
+cartesia/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cartesia/utils/deprecated.py,sha256=2cXvGtrxhPeUZA5LWy2n_U5OFLDv7SHeFtzqhjSJGyk,1674
+cartesia/utils/retry.py,sha256=O6fyVWpH9Su8c0Fwupl57xMt6JrwJ52txBwP3faUL7k,3339
+cartesia/utils/tts.py,sha256=TbvBZqHR6LxPim6s5RyGiURi4hIfqWt3KUk5QYOOhfc,2177
+cartesia-1.3.0.dist-info/METADATA,sha256=eedG5B4V6MxvDuPUMdYwp6UHrX6yQ6dJTMRRZxq1-UA,20976
+cartesia-1.3.0.dist-info/WHEEL,sha256=C2FUgwZgiLbznR-k0b_5k3Ai_1aASOXDss3lzCUsUug,87
+cartesia-1.3.0.dist-info/licenses/LICENSE.md,sha256=PT2YG5wEtEX1TNDn5sXkUXqbn-neyr7cZenTxd40ql4,1074
+cartesia-1.3.0.dist-info/RECORD,,

{cartesia-1.1.0.dev0.dist-info → cartesia-1.3.0.dist-info}/WHEEL

@@ -1,5 +1,4 @@
 Wheel-Version: 1.0
-Generator: setuptools (75.2.0)
+Generator: hatchling 1.26.3
 Root-Is-Purelib: true
 Tag: py3-none-any
-

cartesia-1.1.0.dev0.dist-info/RECORD

@@ -1,24 +0,0 @@
-cartesia/__init__.py,sha256=rS7jIg4iqT0VgnwjzYK25JXxnF5hjZGE_-PGynAqHFo,126
-cartesia/_async_sse.py,sha256=76oIvstzVcWZCbcD8Ps419k1FEHF6lOB5qoHwawvj9k,3327
-cartesia/_async_websocket.py,sha256=Gy0nK3g2HKIBwh-PP1AunEBj83kgFpTGCvrq6tnwg9c,12515
-cartesia/_constants.py,sha256=lquaYIg7IThdmC1fCklnWC8EM7stbSeVCDwRqCzPq-U,389
-cartesia/_logger.py,sha256=vU7QiGSy_AJuJFmClUocqIJ-Ltku_8C24ZU8L6fLJR0,53
-cartesia/_sse.py,sha256=CugabGUAUM-N2BruxNFxDB20HyxDlRdbN-J_yAzvBMY,5667
-cartesia/_types.py,sha256=2fTSCwjL9lJ3jsdbs0P9fHsjkhejyrrYt6oqIXGk1y4,4488
-cartesia/_websocket.py,sha256=CpqkShdl4qBjCGMR8s6dEBHK0LJxkrG-FjbPLhjOP-U,14735
-cartesia/async_client.py,sha256=y_K_Yuv0weA4k9ZYD0M9bNM3x3frsq07tqkg7R9h0-o,2714
-cartesia/async_tts.py,sha256=IbWVRKklNClXASR6ylHaukcMRR304LUguqc4yMopbDU,2076
-cartesia/client.py,sha256=OS1ORUSlR8Jg-em1imeTAFfwkC85AQFnw8PYtTdUuC8,2364
-cartesia/resource.py,sha256=wpnB3IPcTdxYSp0vxSkpntp4NSvqvnwUWF-0ZpgWV9o,1585
-cartesia/tts.py,sha256=RiADE9wjukfq595DrtgBZY8OKoTaFBzef0wCG93yvFM,5345
-cartesia/version.py,sha256=7mgCxmrH_BGK1I131E6uUNAUW91B4b9YhDrULqohQi0,27
-cartesia/voices.py,sha256=DB4tEiSJp7jfnQM0HoiSFS09ZY2oAFbOwMlKe6pofTs,5606
-cartesia/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-cartesia/utils/deprecated.py,sha256=2cXvGtrxhPeUZA5LWy2n_U5OFLDv7SHeFtzqhjSJGyk,1674
-cartesia/utils/retry.py,sha256=O6fyVWpH9Su8c0Fwupl57xMt6JrwJ52txBwP3faUL7k,3339
-cartesia/utils/tts.py,sha256=7tJmdyOYwe2QIav5d1UZxhpbcHaYqf7A77bBOlb4U_g,2100
-cartesia-1.1.0.dev0.dist-info/LICENSE.md,sha256=PT2YG5wEtEX1TNDn5sXkUXqbn-neyr7cZenTxd40ql4,1074
-cartesia-1.1.0.dev0.dist-info/METADATA,sha256=TORvxC_g0KM3OdvEpJaGSEwJ2ZawfMM_m4J58MI8ZYU,21190
-cartesia-1.1.0.dev0.dist-info/WHEEL,sha256=OVMc5UfuAQiSplgO0_WdW7vXVGAt9Hdd6qtN4HotdyA,91
-cartesia-1.1.0.dev0.dist-info/top_level.txt,sha256=rTX4HnnCegMxl1FK9czpVC7GAvf3SwDzPG65qP-BS4w,9
-cartesia-1.1.0.dev0.dist-info/RECORD,,

cartesia-1.1.0.dev0.dist-info/top_level.txt

@@ -1 +0,0 @@
-cartesia