1. Fix Multi/Infinite Talk long video gen bug.
2. Ignore using Replace Video Latent Frames node in the 1st run for Cuts workflows.
3. Update SaveStringListToJSON_JK node, save as list by default instead of string.
4. Add Wan Move ff2v video gen sub workflow.
5. Update FlashVSR video upscale sub workflow using offload loading device.
6. Add Video MaMa for mask refinement to video auto mask sub workflow.
1. Add audio selections to Random Prompter.
2. Update workflows with Random Prompter.
3. Add Trellis multiview mesh generation workflow and replace remesh node with remesh with Quad node.
4. Add LTXV2 video extend sub workflow to it's main workflow.
5. Update CutAudio_JK node, generate silent audio if no audio input.
1. Update the JK_module_Imgen_Seamless Texture workflow and remove deprecated nodes.
2. Update the JK_workflow_imgen_PortraitMaster+RandomPrompter workflow and using the latest Portrait Master nodes.
1. Add LTXV2 video generation workflow, supports t2v | ff2v | lf2v | flf2v | frames2v | v2v, supports ControlNet | upscale.
2. Add more camera motion to Random Prompter.
3. Change SeamlessTile to "Modify in place" for all SD15 | SDXL workflows | Group Nodes | Subgraphs.
1. Add Pose Control workflows for WAN Animate | OTA | Steady Dance | SCAIL.
2. Add | Update Wan (long) video generation with pose control workflows including Vace, Wan Animate, SCAIL, One to All, Sdeady Dance.
3. Replace all Enchance And Resize Hint Images JK nodes in video gen workflows with KJ Resize v2 node.
4. Since KJ make Uni3C suport t2v model, which is greate, I add Uni3C to VACE and t2v workflows.
5. Change default attention to sage attention instead of radial attention.
6. Clip Loader Multi GPU node doesn't work for the latest ComfyUI, it looks like the generation never follows the prompts, especially for t2v. Replace it with the CLIPLoaderDisTorch2MultiGPU node.
1. Add adv3d_viewer_jk features:
① wireframe, Normal, Depth, lineart, canny, edge, contour, ssao, and gtao shading support.
② shading and default light gui support.
2. Update img2mesh 3D Data workflow.
1. Add Wan 2.2 FFGO ref2v sub workflows.
2. Add Stable-Video-Infinity lora supports to Wan 2.1 long | Cuts ff2v sub workflows.
3. Add Wan 2.1 Ditto video edit sub workflows.
4. Add Wan 2.1 longcat refiner v2v sub workflows.
1. Add Wan 2.1 lynx ref2v sub-workflow.
2. Add Wan 2.1 MoCha v2v subject replacement sub-workflow.
3. Add Wan 2.1 FlashVSR video upscale sub-workflow.
4. Add Wan 2.1 UniLumos v2v relight sub-workflow.
5. Add Wan 2.1 LongCat long video gen sub-workflow.
6. Add Wan 2.2 Time-to-Move ff2v sub-workflow.
7. Add Wan 2.2 5B Ovi txt2v and ff2v sub-workflow.
8. Add Wan 2.2 5B Lucy Edit v2v sub-workflow.
1. Add System Prompter node for generating single image | shot script system prompt for LLM/VLM model. Supports JSON format and multi-language output request. QWen3-VL is recommended.
2. Add QWen3-VL to auto-prompt workflows along with System Prompter node.
1. Add HuMo t2v | ref2v + a2v | infinitetalk workflows.
2. Add a description about the fp8_e5m2 model, which is available for Triton.
3. Add Audio Cuts JK🐉 | Cut Audio JK🐉 | Cut Audio Index JK🐉 | Cut Audio Cuts JK🐉 | Cut Audio Loop JK🐉 | Get Nth String JK🐉 | Wan Frame Count JK🐉 nodes.
4. Add first two of Wan Video "Cuts" series workflows (VACE | Talk), supports multiple video generation based on scene and audio cuts.
1. Fix Wan Model Loader vace model slot renaming issue in all the latest workflows using Subgraph.
2. Add Wan t2v MTVCrafter + multitalk | infinite talk workflow. Add Wan ff2v MTVCrafter (+ fantasy portrait) + multitalk | infinite talk workflow.
3. Add Wan VACE keyframe2v long video gen workflow.
1. Fix error issue in all latest workflows using Subgraph: KJ Get node | rgthree Seed node will cause an error if it is in bypass state and is linked to a Subgraph.
2. Add Wan t2v MTVCrafter workflow. Add Wan ff2v MTVCrafter (+Uni3C) workflow.
1. Add conditioning combine and Vace to Video nodes for Wan Phatom workflows.
2. Load String List From JSON JK🐉 node support both string list and JSON object.
3. Add Wan Wrapper Multitalk + Vace workflow.
4. Share Multitalk and Infinite talk workflows.
5. Add Wan t2v Fantasy Talking | Multitalk | Infinite talk workflows.
6. Add Wan wrapper ff2v Fantasy Portrait (+ Multitalk | Infinite talk) workflow.
1. Create Subgraphs based on my Group Nodes and recreate all workflows using Subgraph.
2. Move all latest workflows using Group Node to subfolder named `ComfyUI_~frontend-1.23.4` or `Krita_~frontend-1.23.4_v1.19.20`.
3. Add LoRA Stack Model Only JK🐉, Apply LoRA Stack Model Only JK🐉.
4. Add Set LoRA | NAG | Set Radial Attention | FreeInit features to all WAN video gen workflows.
5. Add native VACE ff2v | VACE flf2v | VACE v2v | ATI ff2v | Fun control camera ff2v | Skyreel A2 ref2v | Phantom ref2v workflows.
6. Add native and wrapper Wan MAGREF workflows.
7. Update ref2v (subject2v) Wan workflows.
8. Add MultiTalk ff2v and v2v Wan Wrapper workflows.
9. Add Multitalk | Fantasy Talking v2v lip sync Wan Wrapper workflows.
10. Add Uni3C Camera Control to ff2v Wan Wrapper workflows.
11. Add MiniMax Remover v2v Wan Wrapper workflow.
12. Add Stand-In ref2v Wan wrapper workflows.
13. Add VACE | Phantom | Stand-In ref+v2v Face Swap workflows.
14. Add Create Loop Schedule List JK🐉 node.
15. Update SkyreelV2 DF long video Wan Wrapper workflow, using for loop.
16. Use the Torch Compile Model Wan Video v2 node for Native Wan Video workflows.
17. Set minimum resolution of Crop by Mask to 64.
18. Refine Common Group Node issue fixed.
19. Add Ksampler Adv Parameters Default JK🐉 node and 🐉 JK:::KSampler (High+Low) Group Node for Wan2.2.
20. Fix Switch(Any) node output datatype error in all workflows.
21. Update Refine Common Group Node for all workflows (ComfyUI frontend 1.23.4 or earlier).
1. Find a workaround to recreate all of my original Group Nodes for `comfyui-frontend-package>=1.16.0`.
2. Seperate all workflows into two versions: `comfyui-frontend-package<=1.15.13` & `comfyui-frontend-package>=1.16.0`. Workflows for `comfyui-frontend-package<=1.15.13` are moved to the subfolder with a suffix name `~frontend-1.15.13`.
3. API workflows are separated into two versions: one for `ComfyUI-JakeUpgrade v1.9.16 or earlier`, the other for `ComfyUI-JakeUpgrade v1.9.17 or later`. API Workflows for `ComfyUI-JakeUpgrade v1.9.16 or earlier` are moved to the subfolder with a suffix name `~v1.9.16.
4. Cleanup and rearrange the codes, and press Deprecated and Experimental to distinguish.
5. Remove java script.
6. Re-write Lora | ControlNet | Embedding nodes without java script control.
1. Add Dilated ControlNet for Wan Video t2v workflows.
2. Add Canny & HED Control Video generation methods for Dilated ControlNet.
3. Add ATI Trajectory generation method to Video-Auto Motion module.
4. Add Save|Load StringListToJSON nodes to save|load ATI Trajectory presets.
5. Use bf16 Wan Video VAE by default for all Wan Video generation workflows.
6. Add Multi-GPU version of Flux Image Generation workflow.
1. Use 10 camera for MV Sampling by default.
2. Add Mesh MV sample generation module workflow using IG2MV, MV Adapter and Stable X Delight.
3. Add Multi-GPU version of Wan Video Generation workflows.
4. Add the Wan Vace fp16 model to the model list instruction.
1. Add Trim Video Latent Node ot Native Wan VACE workflow. Update Native t2v and r2v Wan VACE workflow.
2. Remove Save Image and Save Webm nodes for Auto Mask | Auto Motion | Control Video Module workflow. Video Combine can save as both.
1. Add Florence 2 to Auto Prompt Workflow.
2. Add Segment Anything 2 to Detailer Workflow.
3. Add Flux Fill to Flux imgen group node and workflow.
4. Add RoughOutline_JK node to generate motion outline.
5. Add OpenDWPose_JK node to combine openpose and dwpose.
6. Add Wan Video main workflow and module workflows.
1. Add Stop At Clip Layer to image refine group nodes.
2. Add Flux Guidance to common imgen | refine group nodes.
3. Add ComfyUI API workflows: controlnet preprocessor | upscale | refine.
1. Add Inject Latent Noise to advanced imgen group nodes.
2. Fix the Input Mask logic of advanced imgen group nodes.
3. Add "SD3 Prompts switch" for advanced SD3 imgen group nodes.
4. Switch clip l and g bug fix.
5. Follow the update of Inspire-Pack Random Noise node for imgen group nodes.
6. Image Generation | API | Kirta workflows updated.
1. ControlNet image supports alpha. Add A1111-ControlNet-like effective mask for ControlNet (Stack) Apply nodes.
2. Load Image With Alpha, Make Image Grid, and Split Image Grid nodes added.
3. IF AI nodes recreated.
4. ComfyUI API workflow added.
5. Get Mesh 3D data workflow added.
1. Add SAI API Replace Background and Relight. Add missing parameters of SAI API nodes.
2. Add Detail Daemon Custom Nodes to most of image|mesh generation workflows and group nodes.
3. Add SDXL Target Res JK🐉 node to fix SDXL Text Encode Target Resolution not working.
4. Add Image | Latent Crop by Mask, Resize, Crop by Mask and Resize, Stitch nodes. Add Crop and Stitch operation for Image Gen and Inpaint Group Nodes.
5. Update img2mesh MV upscale method to achieve better results: SD15 img2img + SDXL Refine + Ultimate Upscale.
6. Add Bool_Or and Bool_And nodes for workflows need to be group. Replace Bool Binary Operation node (Or as default) with Bool_Or.
7. Add Color Grading Node.
8. Seperate Multi-ControlNet Param Stack node to ControlNet Loader and Multi-ControlNet Param Stack.
9. Add "None" selection as default to IPAdapter Unified Loader for better workflow control, especially in API mode.
10. Add Image Refine Group Node.
11. Krita image generation workflows updated.
12. Add Krita Refine, Upscale and Refine, Hand fix, CN preprocessor, remove bg and SAI API module series.
13. Remove JK🐉::Pad Image for Outpainting. Remove JK🐉::CLIPSegMask group node, replace it with Essentials CLIPSeg node.
14. Remove [WAS Node Suite] and [Layer Style] Custom Node from must-have-list for their complexity.
15. Remove useless Group Nodes: JK🐉::Image to Prompt (LLava Local), JK🐉::Image to Prompt Advanced (LLava Local), JK🐉::SegAnythingMask, JK🐉::txt2img.
16. SD3 | Flux image generation workflow | group nodes support Stop At Clip Layer (needs ComfyUI v0.3.8up).
jakechai