[New Model] Support IndexTTS 2.5 (#5957)
Signed-off-by: Sysy <1370724210@qq.com> Co-authored-by: Yueqian Lin <70319226+linyueqian@users.noreply.github.com>
This commit is contained in:
1 parent
1ff2753469
commit
58007c0084
77 files changed
+7123
-519
No files matched your search
@@ -743,6 +743,48 @@ class TestTTSMethods:
|
||||
assert speech_server._is_tts is False
|
||||
assert speech_server._tts_stage is None
|
||||
|
||||
@pytest.mark.parametrize(
|
||||
("model_type", "expected_speed"),
|
||||
[("indextts2_5", 1.0), ("qwen3_tts", 2.0)],
|
||||
)
|
||||
def test_audio_encode_speed_respects_adapter_native_control(
|
||||
self,
|
||||
speech_server,
|
||||
model_type,
|
||||
expected_speed,
|
||||
):
|
||||
speech_server._tts_model_type = model_type
|
||||
request = OpenAICreateSpeechRequest(input="Hello", speed=2.0)
|
||||
|
||||
assert speech_server._audio_encode_speed(request) == expected_speed
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_non_streaming_native_speed_skips_generic_audio_adjustment(
|
||||
self,
|
||||
speech_server,
|
||||
mocker: MockerFixture,
|
||||
):
|
||||
async def mock_generate():
|
||||
yield create_mock_audio_output_for_test()
|
||||
|
||||
speech_server._tts_model_type = "indextts2_5"
|
||||
mocker.patch.object(
|
||||
speech_server,
|
||||
"_prepare_speech_generation",
|
||||
new=mocker.AsyncMock(return_value=("speech-native-speed", mock_generate(), {})),
|
||||
)
|
||||
create_audio = mocker.patch.object(
|
||||
speech_server,
|
||||
"create_audio",
|
||||
return_value=SimpleNamespace(audio_data=b"dummy", media_type="audio/wav"),
|
||||
)
|
||||
|
||||
await speech_server._generate_audio_bytes(OpenAICreateSpeechRequest(input="Hello", speed=2.0))
|
||||
|
||||
audio_obj = create_audio.call_args.args[0]
|
||||
assert isinstance(audio_obj, CreateAudio)
|
||||
assert audio_obj.speed == 1.0
|
||||
|
||||
def test_is_tts_detection_with_tts_stage(self, mocker: MockerFixture):
|
||||
"""Test TTS model detection when TTS stage exists."""
|
||||
mock_engine_client = mocker.MagicMock()
|
||||
@@ -2178,17 +2220,17 @@ class TestStreamingProtocolValidation:
|
||||
assert req.is_streaming() is False
|
||||
|
||||
def test_stream_validation_errors(self):
|
||||
"""stream=True requires response_format in ('pcm', 'wav') and speed=1.0."""
|
||||
"""The request schema validates formats; model-aware speed checks happen in serving."""
|
||||
with pytest.raises(ValidationError, match="requires response_format='pcm' or 'wav'"):
|
||||
OpenAICreateSpeechRequest(input="Hello", stream=True, response_format="mp3")
|
||||
with pytest.raises(ValidationError, match="Speed adjustment is not supported"):
|
||||
OpenAICreateSpeechRequest(input="Hello", stream=True, response_format="pcm", speed=2.0)
|
||||
request = OpenAICreateSpeechRequest(input="Hello", stream=True, response_format="pcm", speed=2.0)
|
||||
assert request.speed == 2.0
|
||||
|
||||
def test_stream_format_audio_validation_errors(self):
|
||||
with pytest.raises(ValidationError, match="requires response_format='pcm' or 'wav'"):
|
||||
OpenAICreateSpeechRequest(input="Hello", stream_format="audio", response_format="mp3")
|
||||
with pytest.raises(ValidationError, match="Speed adjustment is not supported"):
|
||||
OpenAICreateSpeechRequest(input="Hello", stream_format="audio", response_format="pcm", speed=2.0)
|
||||
request = OpenAICreateSpeechRequest(input="Hello", stream_format="audio", response_format="pcm", speed=2.0)
|
||||
assert request.speed == 2.0
|
||||
|
||||
def test_stream_valid(self):
|
||||
"""stream=True + response_format in ('pcm', 'wav') + speed=1.0 is accepted as SSE."""
|
||||
@@ -2271,6 +2313,18 @@ class TestStreamingResponse:
|
||||
models=mock_models,
|
||||
request_logger=mocker.MagicMock(),
|
||||
)
|
||||
speech_server._tts_model_type = None
|
||||
speech_server._adapter = None
|
||||
speech_server._diffusion_mode = False
|
||||
mocker.patch.object(speech_server, "_uses_native_speed_control", return_value=False)
|
||||
mocker.patch.object(
|
||||
speech_server,
|
||||
"create_error_response",
|
||||
side_effect=lambda message, **_: JSONResponse(
|
||||
status_code=400,
|
||||
content={"error": {"message": message}},
|
||||
),
|
||||
)
|
||||
|
||||
original_create_speech = speech_server.create_speech
|
||||
sig = signature(original_create_speech)
|
||||
@@ -2285,6 +2339,7 @@ class TestStreamingResponse:
|
||||
|
||||
app = FastAPI()
|
||||
app.add_api_route("/v1/audio/speech", speech_server.create_speech, methods=["POST"], response_model=None)
|
||||
app.state.speech_server = speech_server
|
||||
return app
|
||||
|
||||
@staticmethod
|
||||
@@ -2393,6 +2448,39 @@ class TestStreamingResponse:
|
||||
assert response.status_code in (400, 422)
|
||||
assert "audio/" not in response.headers.get("content-type", "")
|
||||
|
||||
@pytest.mark.parametrize("stream_format", ["sse", "audio"])
|
||||
def test_native_speed_control_accepts_streaming_speed(
|
||||
self,
|
||||
streaming_app,
|
||||
mocker: MockerFixture,
|
||||
stream_format: str,
|
||||
):
|
||||
speech_server = streaming_app.state.speech_server
|
||||
speech_server._tts_model_type = "indextts2_5"
|
||||
speech_server._uses_native_speed_control.return_value = True
|
||||
|
||||
async def prepare(request, request_id=None, **kwargs):
|
||||
del kwargs
|
||||
generator = speech_server.engine_client.generate(prompt={}, request_id=request_id)
|
||||
return request_id, generator, {"duration_factor": [1.0 / request.speed]}
|
||||
|
||||
mocker.patch.object(speech_server, "_prepare_speech_generation", side_effect=prepare)
|
||||
|
||||
client = TestClient(streaming_app)
|
||||
response = client.post(
|
||||
"/v1/audio/speech",
|
||||
json={
|
||||
"input": "Hello",
|
||||
"stream_format": stream_format,
|
||||
"response_format": "pcm",
|
||||
"speed": 2.0,
|
||||
},
|
||||
)
|
||||
|
||||
assert response.status_code == 200
|
||||
expected_media_type = "text/event-stream" if stream_format == "sse" else "audio/pcm"
|
||||
assert expected_media_type in response.headers["content-type"]
|
||||
|
||||
@pytest.fixture
|
||||
def erroring_streaming_app(self, mocker: MockerFixture):
|
||||
"""Test app whose mock engine raises mid-stream, to exercise the SSE error event."""
|
||||
@@ -2616,6 +2704,18 @@ def test_streaming_speech_session_config_accepts_non_streaming_mode():
|
||||
assert config.non_streaming_mode is True
|
||||
|
||||
|
||||
def test_streaming_speech_session_config_scopes_native_speed_to_http():
|
||||
with pytest.raises(
|
||||
ValidationError,
|
||||
match="native speed control is only available through HTTP streaming",
|
||||
):
|
||||
StreamingSpeechSessionConfig(
|
||||
stream_audio=True,
|
||||
response_format="pcm",
|
||||
speed=1.5,
|
||||
)
|
||||
|
||||
|
||||
class TestAsyncOmniSupportedTasks:
|
||||
"""Test that AsyncOmni reports correct supported tasks based on output modalities."""
|
||||
|
||||
|
||||
Reference in new issue
Block a user