feat: support speech to speech translation in Gemini Live

PiperOrigin-RevId: 914523879

Author: yinghsienwu

CHANGELOG.md

@@ -1,12 +1,5 @@
 # Changelog
 
-## [2.2.0](https://github.com/googleapis/python-genai/compare/v2.1.0...v2.2.0) (2026-05-12)
-
-
-### Features
-
-* Added missing FunctionCallResultDelta type and `arguments` field to the ArgumentDelta type ([b0131f8](https://github.com/googleapis/python-genai/commit/b0131f80c2da97926bcef00b4abef9334272ee7a))
-
 ## [2.1.0](https://github.com/googleapis/python-genai/compare/v2.0.1...v2.1.0) (2026-05-12)
 
 

google/genai/_live_converters.py

@@ -933,6 +933,13 @@ def _LiveConnectConfig_to_mldev(
         ],
     )
 
+  if getv(from_object, ['stream_translation_config']) is not None:
+    setv(
+        parent_object,
+        ['setup', 'generationConfig', 'streamTranslationConfig'],
+        getv(from_object, ['stream_translation_config']),
+    )
+
   return to_object
 
 
@@ -1113,6 +1120,12 @@ def _LiveConnectConfig_to_vertex(
         [item for item in getv(from_object, ['safety_settings'])],
     )
 
+  if getv(from_object, ['stream_translation_config']) is not None:
+    raise ValueError(
+        'stream_translation_config parameter is only supported in Gemini'
+        ' Developer API mode, not in Gemini Enterprise Agent Platform mode.'
+    )
+
   return to_object
 
 

google/genai/_tokens_converters.py

@@ -474,6 +474,13 @@ def _LiveConnectConfig_to_mldev(
         ],
     )
 
+  if getv(from_object, ['stream_translation_config']) is not None:
+    setv(
+        parent_object,
+        ['setup', 'generationConfig', 'streamTranslationConfig'],
+        getv(from_object, ['stream_translation_config']),
+    )
+
   return to_object
 
 

google/genai/tests/live/test_live.py

@@ -1610,6 +1610,55 @@ async def test_bidi_setup_to_api_with_transparent_session_resumption(vertexai):
   assert result == expected_result
 
 
+@pytest.mark.parametrize('vertexai', [True, False])
+@pytest.mark.asyncio
+async def test_bidi_setup_to_api_with_stream_translation_config(vertexai):
+  api_client = mock_api_client(vertexai=vertexai)
+
+  # Test 1: Config defined using dict representation.
+  config_dict = {
+      'stream_translation_config': {
+          'echo_target_language': True,
+          'target_language_code': 'es',
+      },
+  }
+
+  with pytest_helper.exception_if_vertex(api_client, ValueError):
+    result = await get_connect_message(
+        api_client=api_client, model='test_model', config=config_dict
+    )
+
+  if not vertexai:
+    expected_result = {
+        'setup': {
+            'model': 'models/test_model',
+            'generationConfig': {
+                'streamTranslationConfig': {
+                    'echo_target_language': True,
+                    'target_language_code': 'es',
+                },
+            },
+        }
+    }
+    assert result == expected_result
+
+  # Test 2: Config defined using types.LiveConnectConfig.
+  config = types.LiveConnectConfig(
+      stream_translation_config=types.StreamTranslationConfig(
+          echo_target_language=True,
+          target_language_code='es',
+      )
+  )
+
+  with pytest_helper.exception_if_vertex(api_client, ValueError):
+    result = await get_connect_message(
+        api_client=api_client, model='test_model', config=config
+    )
+
+  if not vertexai:
+    assert result == expected_result
+
+
 @pytest.mark.parametrize('vertexai', [True, False])
 def test_parse_client_message_str( mock_websocket, vertexai):
   session = live.AsyncSession(

google/genai/types.py

@@ -19507,6 +19507,40 @@ class RealtimeInputConfigDict(TypedDict, total=False):
 RealtimeInputConfigOrDict = Union[RealtimeInputConfig, RealtimeInputConfigDict]
 
 
+class StreamTranslationConfig(_common.BaseModel):
+  """Config for stream translation."""
+
+  echo_target_language: Optional[bool] = Field(
+      default=None,
+      description="""If true, the model will generate audio when the target language is
+      spoken, essentially it will parrot the input. If false, we will not produce
+      audio for the target language.""",
+  )
+  target_language_code: Optional[str] = Field(
+      default=None,
+      description="""The target language for translation. Supported values are BCP-47
+      language codes (e.g. "en", "es", "fr").""",
+  )
+
+
+class StreamTranslationConfigDict(TypedDict, total=False):
+  """Config for stream translation."""
+
+  echo_target_language: Optional[bool]
+  """If true, the model will generate audio when the target language is
+      spoken, essentially it will parrot the input. If false, we will not produce
+      audio for the target language."""
+
+  target_language_code: Optional[str]
+  """The target language for translation. Supported values are BCP-47
+      language codes (e.g. "en", "es", "fr")."""
+
+
+StreamTranslationConfigOrDict = Union[
+    StreamTranslationConfig, StreamTranslationConfigDict
+]
+
+
 class LiveConnectConfig(_common.BaseModel):
   """Session config for the API connection."""
 
@@ -19645,6 +19679,9 @@ class LiveConnectConfig(_common.BaseModel):
       response.
       """,
   )
+  stream_translation_config: Optional[StreamTranslationConfig] = Field(
+      default=None, description="""Config for stream translation."""
+  )
 
 
 class LiveConnectConfigDict(TypedDict, total=False):
@@ -19763,6 +19800,9 @@ class LiveConnectConfigDict(TypedDict, total=False):
       response.
       """
 
+  stream_translation_config: Optional[StreamTranslationConfigDict]
+  """Config for stream translation."""
+
 
 LiveConnectConfigOrDict = Union[LiveConnectConfig, LiveConnectConfigDict]
 

google/genai/version.py

@@ -13,4 +13,4 @@
 # limitations under the License.
 #
 
-__version__ = '2.2.0' # x-release-please-version
+__version__ = '2.1.0' # x-release-please-version

pyproject.toml

@@ -3,7 +3,7 @@ requires = ["setuptools", "wheel", "twine>=6.1.0", "packaging>=24.2", "pkginfo>=
 
 [project]
 name = "google-genai"
-version = "2.2.0"
+version = "2.1.0"
 description = "GenAI Python SDK"
 readme = "README.md"
 license = "Apache-2.0"