# Core AI stack
diffusers==0.34.0
transformers==4.53.1
tokenizers>=0.20.3
accelerate>=1.1.1
tqdm
imageio
imageio-ffmpeg
einops
sentencepiece
open_clip_torch>=2.29.0

# Video & media
moviepy==1.0.3
av
ffmpeg-python
pygame>=2.1.0
sounddevice>=0.4.0
soundfile
mutagen
pyloudnorm
librosa==0.11.0
speechbrain==1.0.3
 
# UI & interaction
gradio==5.29.0
dashscope
loguru

# Vision & segmentation
opencv-python>=4.9.0.80
segment-anything
rembg[gpu]==2.0.65
onnxruntime-gpu
decord
timm

# Config & orchestration
omegaconf
hydra-core
easydict
pydantic==2.10.6

# Math & modeling
torchdiffeq>=0.2.5
tensordict>=0.6.1
mmgp==3.6.0
peft==0.15.0
matplotlib

# Utilities
ftfy
piexif
pynvml
misaki

# Optional / commented out
# transformers==4.46.3  # for llamallava pre-patch
# rembg==2.0.65         # non-GPU fallback
# huggingface_hub[hf_xet]  # slows down everything
# num2words
# spacy