Update: Base64 support + Local file fallback + TTS fixes

done

Optimize: Fast transport and build-time warming

Production configurations updated

Fix: Push only necessary code and s9.jpg asset

Fix: Push only necessary code and s9.jpg asset

Fix: Push only necessary code and s9.jpg asset

Fix: Enable CPU mode, upload assets, and full LFS tracking

Final fix: all weights and onnxruntime version

Final fix: all weights and onnxruntime version

Fix: Corrected landmarker URL and switched to KwaiVGI official

Fix: Implementation plan - ONNX downgrade and weight links

Fix: KlingTeam paths and upgrade onnxruntime

Fix download URLs in download_weights.py per user request

Fix: Updated to official KwaiVGI links and stable onnxruntime

Fix: Corrected model URLs for KwaiVGI repository

Fix download_weights.py: encapsulated logic in main()

Install g++ and build-essential for insightface compilation

Rename readme.md to README.md (case sensitive for HF)

Fix Dockerfile: Replace libgl1-mesa-glx with libgl1

Fresh deployment to Hugging Face