setuptools<82,>=77.0.3
configargparse>=1.2.1
packaging
typeguard
humanfriendly
scipy>=1.4.1
numpy>=2.0.0
PyYAML>=5.1.2
soundfile>=0.10.2
omniio>=0.1.2
filelock
sentencepiece==0.2.1
librosa>=0.10.2
torch_complex
einops
tqdm>=4.64
espnet_model_zoo
torch!=2.12.0,!=2.12.1,<2.15,>=2.11.0
torchaudio<2.12,>=2.11.0
huggingface_hub

[all]
espnet[train]
espnet[asr]
espnet[asr2]
espnet[tts]
espnet[enh]
espnet[st]
espnet[s2t]
espnet[s2st]
espnet[spk]
espnet[speechlm]
espnet[egs2]
espnet[mcp]
fairscale
transformers
evaluate

[asr]
espnet-ctc-segmentation>=1.7.4.post1
editdistance
opt_einsum
jiwer

[asr2]
editdistance

[demo]
gradio

[dev]
black
flake8>=3.7.8
pytest<8.4.0,>=7.0.0
pytest-cov>=2.7.1
isort

[doc]
sphinx<9.0.0
sphinx-rtd-theme>=0.2.4
myst-parser
sphinx-argparse>=0.2.5
sphinx-markdown-builder
sphinx-markdown-tables

[egs2]
espnet[train]
gdown
resampy
pysptk>=0.2.1
morfessor
youtube_dl
nnmnkwii
museval>=0.2.1
pystoi>=0.2.2
mir-eval>=0.6
fastdtw
nara_wpe>=0.0.5
sacrebleu>=1.5.1
praatio<7,>=6
scikit-learn>=1.0.0

[enh]
ci_sdr
fast-bss-eval==0.1.3

[mcp]
mcp>=2.0
espnet[asr]
espnet[tts]
espnet[enh]

[s2st]
editdistance
espnet-s3prl>=2026.5.14

[s2t]
editdistance

[sds]
pydub
snac
litgpt==0.4.3
openai-whisper
webrtcvad-wheels
gradio
transformers

[speechlm]
lhotse
sortedcontainers
pyarrow
psutil
transformers==5.14.1
huggingface-hub
joblib
polars
duckdb

[speechlm:sys_platform == "linux"]
torchtitan
liger-kernel
flash-linear-attention>=0.2.2
tilelang>=0.1.12

[spk]
asteroid_filterbanks==0.4.0

[st]
editdistance

[test]
espnet[train]
pytest<8.4.0,>=7.0.0
pytest-timeouts>=1.2.1
pytest-pythonpath>=0.7.3
pytest-cov>=2.7.1
hacking>=2.0.0
mock>=2.0.0
pycodestyle
jsondiff>=2.0.0
flake8>=3.7.8
flake8-docstrings>=1.3.1
black
isort
h5py>=2.10.0
gradio
playwright
mcp>=2.0

[train]
lightning
tensorboard>=1.14
torch_optimizer
matplotlib
nltk>=3.4.5
wandb
editdistance
hydra-core
omegaconf
datasets
dask[distributed]
dask_jobqueue

[tts]
pyworld>=0.3.4
pypinyin<=0.44.0
espnet_tts_frontend
espnet-g2p-en>=2.1.0.post1
jamo==0.4.1
jaconv
