What's Changed

• feat: MediaPipe face detection (CORE-235) by @kijai in #14009
• Multi-threaded load of models from disk (big load time speedups & Offload to disk) (CORE-43,CORE-152,CORE-164,CORE-165,CORE-117) by @rattus128 in #13802
• Repo security stuff. by @comfyanonymous in #14019
• feat(openapi): add OAuth 2.1 + RFC 7591 DCR endpoints by @mattmillerai in #14026
• fix(openapi): correct POST /api/assets/import to importPublishedAssets by @mattmillerai in #14027
• openapi: remove cloud-only job_ids query param from GET /api/assets by @mattmillerai in #14016
• Update MediaPipe nodes to standardize with existing code base (CORE-242) by @alexisrolland in #14025
• chore: Dataset nodes clean-up (CORE-237) by @alexisrolland in #14002
• Update MoGe nodes display names, search aliases and descriptions by @alexisrolland in #14030
• fix(lanczos): correct dimension transposition for single-channel tensors (CORE-228) by @edoardocarmignani in #12679
• Fix reshaping lora application by @rattus128 in #14031
• [Partner Nodes] add OpenRouter LLM node by @bigcat88 in #14007
• Update workflow templates to v0.9.82 by @comfyui-wiki in #14034
• [Partner Nodes] add widget for auto upscaling for the ByteDance2Reference node by @bigcat88 in #14032
• Add backport release workflow. by @comfyanonymous in #14038
• Fix SIGPIPE false negative in Backport Release workflow validation by @Kosinkadink in #14041
• Add validation for source branch in backport workflow by @comfyanonymous in #14042
• Update backport workflow to use commit SHA input by @comfyanonymous in #14043
• fe: Consolidate warnings spam by @rattus128 in #13970
• Update Discord invite link in README.md by @comfyanonymous in #14045
• Adding in And, Or, and Not nodes. by @Pauan in #14004
• Move logic nodes into utils category by @alexisrolland in #14033
• [Partner Nodes] add new Rodin2.5 nodes by @bigcat88 in #14051
• openapi: align response declarations with implementation (5 endpoints) by @mattmillerai in #14058
• Delete the source branch after doing the backport. by @comfyanonymous in #14062
• openapi: rename 55 cloud-side operationIds to match runtime (PR A of 3) by @mattmillerai in #14060
• openapi: rename cloud-side response schemas to match runtime (PR D) by @mattmillerai in #14065
• openapi: add enum values + FeedbackRequest schema for cloud cutover (PR E) by @mattmillerai in #14070
• Remove dead code. by @comfyanonymous in #14072
• Bump comfyui-frontend-package to 1.44.19 by @comfy-pr-bot in #14074
• Fix --use-flash-attention ignored when xformers installed. by @comfyanonymous in #14083
• fix: correct description of where compiled FE files live by @robinjhuang in #14013
• openapi: fix invalid BillingStatus schema (object + enum hybrid) by @mattmillerai in #14071
• Aimdo 0.4.5 + Cold marking of pins by @rattus128 in #14088
• cache-ram: lower thresholds by @rattus128 in #14089
• Add colored logs by @Talmaj in #14036
• Save Image advanced node (CORE-32) by @comfyanonymous in #13850
• Update embedded docs to v0.5.1 by @comfyui-wiki in #14101
• Emit hash alongside asset_hash on all Asset responses by @mattmillerai in #13739
• Add new open-source model and built-in tool blueprints by @comfyui-wiki in #13980
• MultiGPU Work Units For Accelerated Sampling (CORE-184) by @Kosinkadink in #7063
• Remove some useless annotations imports. by @comfyanonymous in #14105
• multigpu: use unet_manual_cast for SelectModelDevice compute dtype by @Kosinkadink in #14108
• fix(LTX AV): pass correct timesteps to cross-attention AdaLN modulation by @izorinLightricks in #14097
• feat: Microsoft Lens support (CORE-248) by @kijai in #14077
• Lens: some cleanup by @kijai in #14112
• openapi: document QueueManageResponse body on POST /api/queue by @mattmillerai in #14117
• feat: Support NVIDIA PixelDiT and PiD (CORE-201) by @kijai in #14103
• Closer memory usage factors for PID by @comfyanonymous in #14123
• Add memory usage factor for lens model. by @comfyanonymous in #14124
• fix: Stop LTXVCropGuides leaving stray latent frames when guides share a start position (CORE-206) by @drozbay in #13882
• [Partner Nodes] feat: improve video references uploading for SeeDance 2 by @bigcat88 in #14098
• Update workflow templates to v0.9.85 by @comfyui-wiki in #14134
• [Partner Nodes] feat: add Krea2 Image nodes by @bigcat88 in #14130
• [V3] convert nodes_lt_upsampler nodes to V3 schema by @bigcat88 in #12423
• [Partner Nodes] feat: Beeble SwitchX nodes by @bigcat88 in #14137
• Support context window for PiD and fix lq_latent rounding by @kijai in #14136
• Disable sage attention in stable audio dit and VAE. by @comfyanonymous in #14148
• chore: Update nodes categories by @alexisrolland in #14145
• feat: add camera intrinsics fields to Load3DCamera info by @jtydhr88 in #14143
• Make Load3D model_file optional by adding "none" choice by @jtydhr88 in #13379
• [Partner Nodes] fix: add runtime check for SeeDance2 image inputs by @bigcat88 in #14152
• [Partner Nodes] chore: update the category of the Beeble nodes by @bigcat88 in #14156
• feat: add overwrite/increment to SaveImageTextDataSetToFolderNode by @zeerd in #13215
• Add unreviewed merge detector for SOC 2 compliance by @luke-mino-altherr in #14146
• Bump pyav package to fix some image loading issues. by @comfyanonymous in #14160
• refactor: drop rotation from Load3DCamera by @jtydhr88 in #14159
• Update and pin comfy-kitchen version to 0.2.9 by @comfyanonymous in #14161
• float: use CK stochastic rounding cuda kernel (CORE-200) by @rattus128 in #13971
• Update workflow templates to v0.9.91 by @comfyui-wiki in #14163
• feat: add model_info output to Load3D node by @jtydhr88 in #14144
• [Partner Nodes] fix: removed "beta" models versions from Grok nodes by @bigcat88 in #14170
• Fix background removal mask output shape by @kijai in #14171
• [Partner Nodes] feat: add new nodes for Tripo3D P1 model by @bigcat88 in #14155
• Remove useless code. by @comfyanonymous in #14178
• Speed up anima a bit on nvidia. by @comfyanonymous in #14181
• feat(io): add File3DPLY / File3DSPLAT / File3DSPZ / File3DKSPLAT types by @jtydhr88 in #14185
• Threaded Loader performance fixes / improvements (+ Aimdo 0.4.6) by @rattus128 in #14116
• Revert deprecation of non-dynamic smart memory (CORE-152 (revert)) by @rattus128 in #14183
• feat: add Preview3DAdvanced node by @jtydhr88 in #14175
• Update embedded docs to v0.5.2 by @comfyui-wiki in #14193
• Speed up ernie model by a bit on nvidia and use higher quality rope. by @comfyanonymous in #14192
• fix: enable XPU MultiGPU CFG Split on Windows (CFG > 1) by @savvadesogle in #14191
• feat: Add gaussian splat nodes by @kijai in #14190
• [Partner Nodes] added grok-imagine-video-1.5 model to the GrokVideo node in First Frame mode by @bigcat88 in #14198
• Remove old portable updater migration code. by @comfyanonymous in #14202
• feat: Add TripoSplat support by @kijai in #14210
• [Partner Nodes] feat: add Flux Virtual Try-On and Erase nodes by @bigcat88 in #14207
• Update workflow templates to v0.9.92 by @comfyui-wiki in #14212
• [Partner Nodes] fix: respect VideoSlice trim when resizing videos by @bigcat88 in #14213

New Contributors

• @edoardocarmignani made their first contribution in #12679
• @zeerd made their first contribution in #13215
• @savvadesogle made their first contribution in #14191

Full Changelog: v0.22.0...v0.23.0