- Use librosa instead of torchaudio for mediaStorage
- Created dedicated toggle handlers for Image Inference, Video Inference, Upscaler, Audio Inference, Accelerator Options, Bytedance Provider Settings, and OpenAI Provider Settings
- Fixed Audio Inference widget ordering (use parameters now appear before their corresponding parameters)
- All handlers use the same proven pattern from videoInferenceDimensionsHandler
- Improved widget disabling/enabling functionality across all nodes
- Fix image captioning widget targeting to prevent prompt field overwriting
- Add support for multiple images, custom prompts, and model selection in image captioning
- Add missing audio inference nodes to unified color scheme
- Add missing provider settings nodes to unified color scheme
- Fix tensor boolean evaluation error in image captioning
- Update image captioning to match SDK logic for inputImages/inputImage handling
- Separate prompt input from generated caption output fields
- Added useClipSkip boolean parameter to conditionally include clipSkip in API request
- Added None option to dimensions dropdown to let model determine dimensions automatically
- When None is selected, width and height parameters are not sent to API
- This resolves compatibility issues with models that don't support custom dimensions
- Convert useSteps, useCFGScale, useScheduler, useSeed to BOOLEAN in ImageInference
- Convert useFps to BOOLEAN in VideoInference
- Reorder parameters so 'use' toggles appear above their corresponding parameters
- Fix VideoInference providerSettings input type from DICT to RUNWAREPROVIDERSETTINGS
- Update provider settings handling logic to match ImageInference format
- Improve UX with native boolean toggles and logical parameter ordering
- Add Runware Video Inference node with frame images and reference images support
- Add conditional parameter support for steps and CFG scale in image inference
- Add useFps parameter for video inference with enable/disabled options for Vidu support
Sirshendu Ganguly