cartesia 1.0.10__tar.gz → 1.0.12__tar.gz

{cartesia-1.0.10 → cartesia-1.0.12}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: cartesia
-Version: 1.0.10
+Version: 1.0.12
 Summary: The official Python library for the Cartesia API.
 Home-page: 
 Author: Cartesia, Inc.
@@ -73,6 +73,11 @@ print("The embedding for", voice["name"], "is", voice["embedding"])
 # Clone a voice using filepath
 cloned_voice_embedding = client.voices.clone(filepath="path/to/voice")
 
+# Mix voices together
+mixed_voice_embedding = client.voices.mix(
+    [{ "id": "voice_id_1", "weight": 0.5 }, { "id": "voice_id_2", "weight": 0.25 }, { "id": "voice_id_3", "weight": 0.25 }]
+)
+
 # Create a new voice
 new_voice = client.voices.create(
     name="New Voice",
@@ -504,6 +509,7 @@ You can enhance the voice output by adjusting the `speed` and `emotion` paramete
 
 Speed Options:
 - `slowest`, `slow`, `normal`, `fast`, `fastest`
+- Float values between -1.0 and 1.0, where -1.0 is the slowest speed and 1.0 is the fastest speed.
 
 Emotion Options:
 Use a list of tags in the format `emotion_name:level` where:

{cartesia-1.0.10 → cartesia-1.0.12}/README.md

@@ -56,6 +56,11 @@ print("The embedding for", voice["name"], "is", voice["embedding"])
 # Clone a voice using filepath
 cloned_voice_embedding = client.voices.clone(filepath="path/to/voice")
 
+# Mix voices together
+mixed_voice_embedding = client.voices.mix(
+    [{ "id": "voice_id_1", "weight": 0.5 }, { "id": "voice_id_2", "weight": 0.25 }, { "id": "voice_id_3", "weight": 0.25 }]
+)
+
 # Create a new voice
 new_voice = client.voices.create(
     name="New Voice",
@@ -487,6 +492,7 @@ You can enhance the voice output by adjusting the `speed` and `emotion` paramete
 
 Speed Options:
 - `slowest`, `slow`, `normal`, `fast`, `fastest`
+- Float values between -1.0 and 1.0, where -1.0 is the slowest speed and 1.0 is the fastest speed.
 
 Emotion Options:
 Use a list of tags in the format `emotion_name:level` where:

{cartesia-1.0.10 → cartesia-1.0.12}/cartesia/client.py

@@ -32,7 +32,6 @@ except ImportError:
     IS_WEBSOCKET_SYNC_AVAILABLE = False
 
 from iterators import TimeoutIterator
-from websockets.sync.client import connect
 
 from cartesia._types import (
     DeprecatedOutputFormatMapping,
@@ -261,6 +260,40 @@ class Voices(Resource):
 
         return response.json()
 
+    def mix(self, voices: List[Dict[str, Union[str, float]]]) -> List[float]:
+        """Mix multiple voices together.
+
+        Args:
+            voices: A list of dictionaries, each containing either:
+                        - 'id': The ID of an existing voice
+                        - 'embedding': A voice embedding
+                    AND
+                        - 'weight': The weight of the voice in the mix (0.0 to 1.0)
+
+        Returns:
+            The embedding of the mixed voice as a list of floats.
+
+        Raises:
+            ValueError: If the request fails or if the input is invalid.
+        """
+        url = f"{self._http_url()}/voices/mix"
+
+        if not voices or not isinstance(voices, list):
+            raise ValueError("voices must be a non-empty list")
+
+        response = httpx.post(
+            url,
+            headers=self.headers,
+            json={"voices": voices},
+            timeout=self.timeout,
+        )
+
+        if not response.is_success:
+            raise ValueError(f"Failed to mix voices. Error: {response.text}")
+
+        result = response.json()
+        return result["embedding"]
+
 
 class _TTSContext:
     """Manage a single context over a WebSocket.
@@ -294,6 +327,7 @@ class _TTSContext:
         context_id: Optional[str] = None,
         duration: Optional[int] = None,
         language: Optional[str] = None,
+        add_timestamps: bool = False,
         _experimental_voice_controls: Optional[VoiceControls] = None,
     ) -> Generator[bytes, None, None]:
         """Send audio generation requests to the WebSocket and yield responses.
@@ -307,6 +341,7 @@ class _TTSContext:
             context_id: The context ID to use for the request. If not specified, a random context ID will be generated.
             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.
             _experimental_voice_controls: Experimental voice controls for controlling speed and emotion.
                 Note: This is an experimental feature and may change rapidly in future releases.
 
@@ -341,6 +376,7 @@ class _TTSContext:
             },
             "context_id": self._context_id,
             "language": language,
+            "add_timestamps": add_timestamps,
         }
 
         if duration is not None:
@@ -858,9 +894,9 @@ class TTS(Resource):
             raise ValueError("Only one of voice_id or voice_embedding should be specified.")
 
         if voice_id:
-            voice = {"mode": "id", "id": voice_id}
+            voice = {"id": voice_id}
         else:
-            voice = {"mode": "embedding", "embedding": voice_embedding}
+            voice = {"embedding": voice_embedding}
         if experimental_voice_controls is not None:
             voice["__experimental_controls"] = experimental_voice_controls
         return voice

cartesia-1.0.12/cartesia/version.py

@@ -0,0 +1 @@
+__version__ = "1.0.12"

{cartesia-1.0.10 → cartesia-1.0.12}/cartesia.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: cartesia
-Version: 1.0.10
+Version: 1.0.12
 Summary: The official Python library for the Cartesia API.
 Home-page: 
 Author: Cartesia, Inc.
@@ -73,6 +73,11 @@ print("The embedding for", voice["name"], "is", voice["embedding"])
 # Clone a voice using filepath
 cloned_voice_embedding = client.voices.clone(filepath="path/to/voice")
 
+# Mix voices together
+mixed_voice_embedding = client.voices.mix(
+    [{ "id": "voice_id_1", "weight": 0.5 }, { "id": "voice_id_2", "weight": 0.25 }, { "id": "voice_id_3", "weight": 0.25 }]
+)
+
 # Create a new voice
 new_voice = client.voices.create(
     name="New Voice",
@@ -504,6 +509,7 @@ You can enhance the voice output by adjusting the `speed` and `emotion` paramete
 
 Speed Options:
 - `slowest`, `slow`, `normal`, `fast`, `fastest`
+- Float values between -1.0 and 1.0, where -1.0 is the slowest speed and 1.0 is the fastest speed.
 
 Emotion Options:
 Use a list of tags in the format `emotion_name:level` where:

{cartesia-1.0.10 → cartesia-1.0.12}/tests/test_tts.py

@@ -102,6 +102,11 @@ def test_create_voice(client: Cartesia):
     assert voice["is_public"] is False
     voices = client.voices.list()
     assert voice in voices
+    
+def test_mix_voice(client: Cartesia):
+    logger.info("Testing voices.mix")
+    output = client.voices.mix(voices = [{"id": SAMPLE_VOICE_ID, "weight": 0.1}, {"id": SAMPLE_VOICE_ID, "weight": 0.9}])
+    assert isinstance(output, list)
 
 @pytest.mark.parametrize("stream", [True, False])
 @pytest.mark.parametrize("_experimental_voice_controls", [None, EXPERIMENTAL_VOICE_CONTROLS, EXPERIMENTAL_VOICE_CONTROLS_2])
@@ -190,7 +195,6 @@ def test_websocket_send_timestamps(resources: _Resources, stream: bool):
     
     ws.close()
 
-        
 @pytest.mark.parametrize("_experimental_voice_controls", [None, EXPERIMENTAL_VOICE_CONTROLS, EXPERIMENTAL_VOICE_CONTROLS_2])
 def test_sse_send_context_manager(resources: _Resources, _experimental_voice_controls: VoiceControls):
     logger.info("Testing SSE send context manager")
@@ -456,6 +460,28 @@ def test_sync_continuation_websocket_context_send():
             assert isinstance(out["audio"], bytes)
     finally:
         ws.close()
+        
+def test_sync_context_send_timestamps(resources: _Resources):
+    logger.info("Testing WebSocket send")
+    client = resources.client
+    transcripts = ["Hello, world!", "I'\''m generating audio on Cartesia."]
+
+    ws = client.tts.websocket()
+    ctx = ws.context()
+    output_generate = ctx.send(transcript=chunk_generator(transcripts), voice_id=SAMPLE_VOICE_ID, output_format={
+        "container": "raw",
+        "encoding": "pcm_f32le",
+        "sample_rate": 44100
+    }, model_id=DEFAULT_MODEL_ID, add_timestamps=True)
+
+    has_wordtimestamps = False
+    for out in output_generate:
+        has_wordtimestamps |= "word_timestamps" in out
+        _validate_schema(out)
+
+    assert has_wordtimestamps, "No word timestamps found"
+    
+    ws.close()
     
 @pytest.mark.asyncio
 async def test_continuation_websocket_context_send():

cartesia-1.0.10/cartesia/version.py

@@ -1 +0,0 @@
-__version__ = "1.0.10"