steer-tts / pyproject.toml
mrfakename's picture
Mirror WhissleAI/steer-tts
5bf57f8 verified
Raw History Blame Contribute Delete
1.61 kB
[build-system]
requires = ["setuptools>=68"]
build-backend = "setuptools.build_meta"
[project]
name = "steertts"
version = "1.1.0"
description = "SteerTTS: a small, numerically steerable English TTS (hybrid attention + bidirectional Mamba-3)"
readme = "README.md"
requires-python = ">=3.10"
license = { text = "Code: MIT. Weights: CC BY-NC 4.0. See LICENSE.md" }
# Versions are the ones the release was verified with (fresh env, TEST_LOG.md). torch is a range because the right
# wheel depends on your CUDA / CPU setup; torchaudio (pulled in by vocos) must match the torch you install.
dependencies = [
"torch>=2.4,<2.12",
"numpy>=1.24,<2.5",
"safetensors==0.8.0",
"huggingface_hub>=0.23",
"vocos==0.1.0",
"g2p_en==2.1.0",
"nltk==3.10.3",
"num2words==0.5.14",
"soundfile==0.14.0",
"soxr==1.1.0",
]
[project.optional-dependencies]
prompt = ["transformers==5.19.0"] # free-text directions (MiniLM sentence encoder)
clone = ["transformers==5.19.0"] # zero-shot cloning (WavLM speaker encoder)
measure = ["praat-parselmouth==0.4.7"] # read the 12 controls back from audio
play = ["sounddevice>=0.4"] # `steertts play` / --play (else ffplay or aplay is used)
all = ["transformers==5.19.0", "praat-parselmouth==0.4.7", "sounddevice>=0.4"]
test = ["transformers==5.19.0", "praat-parselmouth==0.4.7", "pytest>=8", "jsonschema>=4.20"]
[project.scripts]
steertts = "steertts.cli:main"
[tool.setuptools]
packages = ["steertts", "steertts.mamba3"]
[tool.setuptools.package-data]
steertts = ["request.schema.json"]
[tool.pytest.ini_options]
testpaths = ["tests"]