tts_gallery / tests /ui_test.py
Michael Hu
fix: keep API parameter names, reject unknown voices, and stop cloned Chatterbox voices reaching other users
4d9637e
Raw History Blame Contribute Delete
2.9 kB
"""Check the UI wiring with fake models, no downloads: python -m tests.ui_test"""
import gradio as gr
from src.ui import create_interface
class FakeTTS:
def __init__(self):
self.calls = []
def get_supported_languages(self):
return ["English", "Chinese"]
def get_available_voices(self, language=None):
return ["v1", "v2"]
def get_default_voice(self):
return "v2"
def generate_speech(self, text, **kwargs):
self.calls.append((text, kwargs))
return "out.wav"
class FakeSTT:
languages = {"English": "en"}
def __init__(self):
self.calls = []
def transcribe(self, audio_path, language=None):
self.calls.append((audio_path, language))
return "heard"
TTS_NAMES = [
"ResembleAI/chatterbox",
"KittenML/KittenTTS",
"KittenML/kitten-tts-2",
"piper-tts",
"hexgrad/kokoro",
]
STT_NAMES = ["SYSTRAN/faster-whisper", "Cactus-Compute/whistle"]
def main():
tts = {name: FakeTTS() for name in TTS_NAMES}
stt = {name: FakeSTT() for name in STT_NAMES}
demo = create_interface(tts, stt)
fns = {f.api_name: f.fn for f in demo.fns.values()}
api = demo.get_api_info()["named_endpoints"]
params = {name: [(p["parameter_name"], p["parameter_has_default"]) for p in ep["parameters"]] for name, ep in api.items()}
assert params["/tts_chatterbox"] == [("text", False), ("language", True), ("audio_prompt", True)], params
assert params["/stt_whistle"] == [("audio", False), ("lang", True)], params
assert not any(name.startswith("/update_voices") for name in api), sorted(api)
expected = {
"tts_chatterbox",
"tts_kittentts",
"tts_kittentts2",
"tts_piper",
"tts_kokoro",
"stt_whisper",
"stt_whistle",
}
assert expected <= set(fns), sorted(fns)
fns["tts_chatterbox"]("hi", "Chinese", "ref.wav")
assert tts["ResembleAI/chatterbox"].calls[-1] == (
"hi",
{"language": "Chinese", "audio_prompt": "ref.wav"},
)
fns["tts_kokoro"]("hi", "English", "v2")
assert tts["hexgrad/kokoro"].calls[-1] == (
"hi",
{"language": "English", "voice": "v2"},
)
fns["tts_kittentts2"]("hi", "v1", "stable", "ref.wav", "")
assert tts["KittenML/kitten-tts-2"].calls[-1][1]["reference_text"] == "heard"
assert stt["SYSTRAN/faster-whisper"].calls[-1] == ("ref.wav", None)
assert fns["stt_whistle"]("clip.wav", "auto") == "heard"
fns["stt_whistle"]("clip.wav", "en")
assert stt["Cactus-Compute/whistle"].calls == [
("clip.wav", None),
("clip.wav", "en"),
]
try:
fns["tts_kittentts"](" ", "v1")
raise AssertionError("empty text must be rejected")
except gr.Error as e:
assert "Enter some text" in str(e), e
print("ui test passed")
if __name__ == "__main__":
main()