Fix: increase generation timeout from 10 to 20 minutes for T4 with CPU offload

Fix: remove invalid --guide_scale arg from generate.py command

Fix: add diffusers, transformers, accelerate, tokenizers to explicit pip install

Fix: move generate.py patch into standalone Python script, remove inline -c

Fix: install ALL deps + patch both wan/__init__.py and generate.py for optional imports

Fix: patch only problematic submodule imports (speech2video, animate)

Fix: use external Python script for patching wan/__init__.py (no shell escaping)

Fix: simple regex-based try/except wrapper for all wan submodule imports

Fix: comprehensive deps + monkey-patch wan/__init__.py to tolerate missing optional modules

Fix: add librosa (transitive dep from speech2video import chain)

Fix: add decord + all Wan2.2 deps for proper import chain

Fix: add missing einops dependency

Fix: install Wan2.2 deps directly (skip flash_attn/dashscope which fail in Docker build)

Fix: download model at startup via huggingface-cli, remove build-time download

Switch to official Wan2.2 inference code for native T2V+I2V support

Fix: enable CPU offload + attention slicing for T4 (14.75 GB VRAM limit)

Fix: total_mem → total_memory for PyTorch 2.8 compatibility

Fix build: add git dependency, install diffusers from source separately

Upgrade UI: gallery, progress tracking, tabs, image preview, polished design

Add Wan2.2 TI2V 5B FastAPI app: text-to-video + image-to-video

initial commit