- Added KlingAI 6@1 (KlingAI 2.5 Turbo Pro) to video models - Fixed frame images to always use base64 data URIs (not UUIDs) - Added convertTensor2IMGBase64Only function for consistent frame image handling - Fixed videoInputsFrame to use correct 'inputImage' key - Updated API endpoint from runware.dev to runware.ai - Ensured both frameImages.py and videoInputsFrame.py use consistent format