requirements.txt

# `diffusers` is installed from the canonical MiniMax-H3 pull request,
# https://github.com/huggingface/diffusers/pull/14371 ("Minimax h3 follow up (review & refactor)"), pinned to a
# **commit** rather than to its `minimax-h3-refactor` branch: the PR is a WIP and its head moves, and this Space's
# blocks subclass its block classes. Re-pin — and re-check `h3_split_blocks.py` against the block names of the new
# head — whenever the PR updates.
#
#   665f578278365ea4a3318cb8c9b66ce6c01204b9 = refs/pull/14371/head at the time of this deploy
--extra-index-url https://download.pytorch.org/whl/cu130
diffusers @ git+https://github.com/huggingface/diffusers.git@665f578278365ea4a3318cb8c9b66ce6c01204b9
torch==2.11.0
torchvision==0.26.0
# A reference soundtrack that is not already at the audio VAE's 32 kHz is resampled with `torchaudio`; a 32 kHz one
# skips the resample entirely, so this is easy to miss.
torchaudio==2.11.0
# The Qwen3-VL processor decides the vision patch count, so a different minor changes the conditioning.
transformers==5.8.0
accelerate==1.14.0
# diffusers pins <2.
huggingface-hub==1.24.0
# No `gradio` or `spaces` pin here on purpose. Hugging Face appends its own
# `gradio[oauth,mcp]==<sdk_version>` and `spaces==<current>` to the install command, and a
# second pin for either one is a different version specifier for the same package, which
# pip refuses to resolve: "Cannot install spaces==0.51.1 and spaces==0.51.3". The gradio
# version comes from `sdk_version` in README.md instead.
# No `kernels` pin on purpose: the Hub attention backends want `kernels>=0.12.3`, and that version breaks
# transformers 5.8.0 at import.
# PyAV decodes the reference media (`MiniMaxH3VideoReference.from_file`).
av
pillow
peft
numpy
requests
safetensors>=0.8.0

# CPU identity correction and atomic per-file LoRA preparation.
opencv-python-headless>=4.10,<5
filelock