fix: add /app/weights dir with write permissions in Dockerfile

fix: correct HF Space URL + add health endpoint + frontend retry/warmup UX

chore: update HF org references from Harshit2907 to ExploWebsite

fix(renderer): use exact plot_videos.py scaling pipeline (* 3 * 240 + offset)

fix(hq): correct config loading, auto-scaling rendering, and dynamic watermarks

fix(hq): correct bos/eos token access in HQ engine

feat: implement automatic standard weight download

fix: grant write permissions for HQ weights directory in Docker

feat: implement HQ AI Bridge mode with high-fidelity motion restoration and HF Hub integration

fix(inference): tune sampling steps to 60 to prevent temporal mode collapse (stillness)

refactor(backend): bulletproof imports for cloud deployment environment

fix(inference): maximize DDIM sampling steps to 100 for peak generative fidelity

refactor(backend): safely load fastapi lifecycles and strictly route absolute imports

Revert: decrease sampling steps to 20 for faster response time as per request

Fix: Increase sampling steps to 90 to prevent low motion variance (still frame issue)

docs: Update README with Quick Run guide and relax PyTorch requirements for Mac M2

feat: integrated cloud-based diffusion inference module