What's Changed

• Bump comfyui-frontend-package to 1.58.5 by @comfy-pr-bot in #17442
• Add better error for when pixel_space VAE wrongly selected. by @comfyanonymous in #17518
• Make SeedVR2 VAE slice length adaptive to free memory by @kijai in #17508
• [Partner Nodes] feat(Pika): add Pika H3 Flex nodes by @bigcat88 in #17509
• [Partner Nodes] feat(Pika): add Pika Soundtrack, SFX and Music nodes by @bigcat88 in #17328
• [Partner Nodes] feat(Tencent): add Tencent HY LLM node by @bigcat88 in #17534
• Turn on local_web_sso and partner_node_token by default by @dante01yoon in #17540
• Update workflow templates to v0.11.102 by @comfyui-wiki in #17541
• Update comfy-kitchen version to 0.2.41 by @comfyanonymous in #17542
• Fix potential crash. by @comfyanonymous in #17543
• Performance improvements for qwen 3.x (Part 4) - Combine Parallel Linear Weights by @rattus128 in #17361
• Name comfy-agent without a version in agent_requirements.txt (CORE-692) by @christian-byrne in #17454
• Fix Remesh Mesh on imported meshes by @kijai in #17531
• Support RelayVSR (CORE-701) by @kijai in #17511
• [Partner Nodes] feat(Kling): add Kling 4.0 Flash model and original audio option to Kling 4.0 nodes by @bigcat88 in #17536
• [Partner Nodes] feat(OpenAI): add GPT Image 2.6 models by @bigcat88 in #17558
• [Partner Nodes] feat(Qwen): move Qwen Image nodes to Comfy Router by @bigcat88 in #17527
• Update workflow templates to v0.11.103 by @comfyui-wiki in #17563
• [Partner Nodes] feat(Gemini): add Omni Flash 1.2 to the Omni video node by @bigcat88 in #17560
• Fix resolution selector megapixel precision by @GACLove in #17555
• Update api-node.md by @alexisrolland in #17569
• Move run_cpu portable script to ./advanced by @comfyanonymous in #17571
• Make the partner node team code owners of comfy_api_nodes by @comfyui-wiki in #17575
• [Partner Nodes] feat(ByteDance): add 4k resolution to Seedance 2.5 by @bigcat88 in #17579
• Move portable fast fp16 script to ./advanced by @comfyanonymous in #17570
• Add video input support to Qwen 3.x by @kijai in #17582
• Add solid mode to Extract Mesh from Splat by @kijai in #17497
• Fix potential database issue. by @comfyanonymous in #17602
• Get Video Components: hold the decoded frames once, not twice by @synap5e in #17562
• Fix crash when using the UNetSelfAttentionMultiply node. by @comfyanonymous in #17603
• [Partner Nodes] feat(OpenAI): add GPT Image V3, GIF and SVG nodes by @bigcat88 in #17591
• [Partner Nodes] chore(Grok): update prices and tooltips for retired image models by @bigcat88 in #17532
• Fix qwen fp8 crashing when generating text. by @comfyanonymous in #17620
• feat: Support WeVisOne-VL (CORE-690) by @kijai in #17632
• Avoid Mixing Prefetched and Non-Prefetched Usage of Combined Linear by @rattus128 in #17625
• Apply repetition/presence penalty in greedy decoding by @kijai in #17635
• feat: Add texture rebaking and support UltraTex (CORE-606) by @kijai in #17636
• Update image editing models in README by @comfyanonymous in #17650
• [Partner Nodes] feat(Breeze): add Breeze TTS 2 voice and text-to-speech nodes by @bigcat88 in #17611
• [Partner Nodes] feat(ByteDance): add Seedream 5.5 model by @bigcat88 in #17631
• feat: Support MiniMax-H3-image (CORE-708) by @kijai in #17655
• Update workflow templates to v0.11.105 by @comfyui-wiki in #17658

New Contributors

• @GACLove made their first contribution in #17555

Full Changelog: v0.44.0...v0.45.0

What's Changed

• [Partner Nodes] feat(Patina): add Patina texture nodes by @bigcat88 in #17384
• [Partner Nodes] feat(VEED): add VEED Dubbing node by @bigcat88 in #17388
• Update workflow templates to v0.11.98 by @comfyui-wiki in #17391
• Switch pylint py version to 3.12 by @comfyanonymous in #17398
• anonymous_file: bypass python for memfd_create by @rattus128 in #17386
• Bump comfyui-frontend-package to 1.58.1 by @comfy-pr-bot in #17375
• Use the cached frontend for pinned versions without a v prefix. by @christian-byrne in #17348
• Look up a pinned frontend release by its tag instead of listing every release. by @christian-byrne in #17402
• Skip concat if only a single cond. by @comfyanonymous in #17413
• [Partner Nodes] feat(BFL): add 1080p, 2K and 4K targets to Flux Video Upscale V2 node by @bigcat88 in #17394
• Support MiniMax-H3 latent upscaler model by @kijai in #17412
• Skip concat if only a single cond in more conds. by @comfyanonymous in #17414
• Support MiniMax Sol-H3-VAE by @kijai in #17411
• [Partner Nodes] feat(WaveSpeed): move FlashVSR and image upscale nodes to Comfy Router by @bigcat88 in #17319
• Update workflow templates to v0.11.100 by @comfyui-wiki in #17432
• Support context windows for MiniMax-H3 by @kijai in #17425
• [Partner Nodes] feat(Ideogram): add context window region editing to Ideogram 4.5 Precise Edit by @bigcat88 in #17428
• Don't call torch.equal when the two tensors are the same. by @comfyanonymous in #17439
• Add a node to set the VAE compute dtype. by @comfyanonymous in #17450
• [Partner Nodes] feat(Pruna): add P-Video Upscale node by @bigcat88 in #17443
• [Partner Nodes] feat(MiniMax): add H3 Max Recast, Extend, Reframe, Insert, Lip Sync and Camera Control nodes by @bigcat88 in #17424
• Stop paging GitHub releases at the first prerelease by @christian-byrne in #17434
• Return execution_meta in the job detail by @christian-byrne in #17397
• Support prores and using different chroma subsampling for video saving. by @comfyanonymous in #17463
• Fix MTPHead unpack crash in prefetch queue cleanup (CORE-644) by @christian-byrne in #17180
• Remove outdated guideline in AGENTS.md by @comfyanonymous in #17482
• Fix fp8_e4m3fn_fast falling back to bf16 compute with dynamic VRAM by @CyrilB1531 in #16941
• [Partner Nodes] feat(Topaz): add Wonder 4 model to Topaz Image Enhance by @bigcat88 in #17491
• Fix VAE Compute Dtype crash on the LTX 2.5 diffusion VAE decoder by @DoronGenzelHass in #17487
• Add TextGenerateV2 with Autogrow image inputs by @kijai in #17492
• Validate chroma subsampling input. by @comfyanonymous in #17501
• [Partner Nodes] feat(Sonilo): add Sonilo Music 1.5 node by @bigcat88 in #17475
• [Partner Nodes] chore(ByteDance): deprecate Seedance 1.x video nodes by @bigcat88 in #17507
• feat: Support PH3Image (CORE-697) by @kijai in #17506
• [Partner Nodes] feat(OpenAI): add GPT-6.1 Terra and Sol models by @bigcat88 in #17493
• Update workflow templates to v0.11.101 by @comfyui-wiki in #17512
• Update embedded docs to v0.5.16 by @comfyui-wiki in #17513
• Turn on local_web_sso and partner_node_token by default by @dante01yoon in #17490
• Revert "Turn on local_web_sso and partner_node_token by default" by @dante01yoon in #17514

New Contributors

• @CyrilB1531 made their first contribution in #16941

Full Changelog: v0.43.0...v0.44.0

What's Changed

• model_prefetch: syncrhonize all graph cleanups by @rattus128 in #17178
• Add rope scaling support to MOSS-image by @kijai in #17172
• [Partner Nodes] Renew partner-node sign-in tokens in API nodes by @dante01yoon in #16943
• [Partner Nodes] feat(HeyGen): add HeyGen Voice instant clone and text-to-speech by @bigcat88 in #17043
• [Partner Nodes] feat(ClarityAI): add Crystal Video Upscaler node by @bigcat88 in #17192
• [Partner Nodes] feat(BFL): add 2K/4K resolutions to FLUX 3 Video nodes by @bigcat88 in #17200
• Support MOSS-image color-depth control by @kijai in #17206
• Add to Readme that MOSS image is supported. by @comfyanonymous in #17212
• [Partner Nodes] feat(BFL): add FLUX 3 Max nodes by @bigcat88 in #17202
• fix: correct top-p sampling threshold by @curealb in #17221
• Update workflow templates to v0.11.92 by @comfyui-wiki in #17234
• [Partner Nodes] feat(Bria): FIBO 2.0 Image Generate/Edit nodes by @bigcat88 in #17229
• Apply single-head LoRAs to MiniMax-H3 PDD head banks by @kijai in #17239
• Update workflow templates to v0.11.93 by @comfyui-wiki in #17238
• Bump comfyui-frontend-package to 1.57.1 by @comfy-pr-bot in #17244
• feat: Add Fibo 2.0 support by @kijai in #17240
• Set minimum Pillow version in requirements. by @comfyanonymous in #17249
• Add Pad Image node with mirror padding and pad to multiple by @DoronGenzelHass in #17224
• Load the optional local agent through comfy-agent's v1 host by @synap5e in #17241
• [Partner Nodes] fix(HeyGen): update HeyGen Voice price badge by @bigcat88 in #17257
• [Partner Nodes] feat(Krea): add ultra thinking mode to Krea 3 Image node by @bigcat88 in #17260
• [Partner Nodes] feat(VEED): add Fabric 1.0, Lipsync 2.0, and other by @bigcat88 in #17242
• [Partner Nodes] feat(Mirelo): add Mirelo SFX nodes by @bigcat88 in #17265
• Update workflow templates to v0.11.94 by @comfyui-wiki in #17268
• Add model integration instructions to AGENTS.md by @comfyanonymous in #17279
• Fix Gemma 4 31B global attention (k_eq_v, 4 global KV heads) by @Nynxz in #17220
• Support MOSS-Image edit LoRAs by @kijai in #17296
• qwen 3.x performance improvements part III - Adaptive tree MTP by @rattus128 in #17273
• Sync offload streams before stealing a pin by @rattus128 in #17297
• Fibo 2.0: Change RoPE scaling to NTK aware by @kijai in #17290
• Fix Fibo2.0 T2I negative prompt handling by @kijai in #17285
• comfy should only contain model related stuff. by @comfyanonymous in #17307
• Comfy-Aimdo 0.6.0 + Dynamic VRAM KV caches part I by @rattus128 in #17274
• [Partner Nodes] feat(Meshy): add Meshy T2.5 Smart Topology tnodes by @bigcat88 in #17280
• [Partner Nodes] feat(PixVerse): add PixVerse V6-Max model by @bigcat88 in #17287
• [Partner Nodes] feat(Bria): add 3 new nodes by @bigcat88 in #17289
• [Partner Nodes] feat(Ideogram): add Ideogram Vector Generate SVG nodes by @bigcat88 in #17298
• Update workflow templates to v0.11.96 by @comfyui-wiki in #17301
• Enable pytorch attention in the VAE on AMD by @0xDELUXA in #17271
• Fuse and strip the Flux2 VAE through comfy-kitchen by @kijai in #17196
• Add ACEScg to Convert Image Color Space and Save Image (Advanced) EXR (CORE-657) by @DoronGenzelHass in #17023
• Add Comfy API console tips by @guill in #17052
• Add attention bias conditioning and image strength for MOSS-Image edit by @kijai in #17320
• Don't use pytorch attention for head dims above 256 on gfx11 by @kijai in #17323
• Update architecture boundaries in AGENTS.md by @comfyanonymous in #17331
• Lower peak MOSS memory a bit. by @comfyanonymous in #17338
• Add deploy to cloud endpoints and feature flag by @guill in #17203
• Add a Branch node by @AustinMroz in #16383
• Add model specific merge node for MOSS. by @comfyanonymous in #17349
• [Partner Nodes] fix(client): respect proxy environment variables by @bigcat88 in #17346
• docs: add Huawei branding to Ascend NPU section in README by @yzeyu71 in #17359
• Bump comfyui-frontend-package to 1.57.4 by @comfy-pr-bot in #17329
• [Partner Nodes] feat(Kling): add Kling 4.0 nodes by @bigcat88 in #17364
• Update embedded docs to v0.5.15 by @comfyui-wiki in #17366
• Update workflow templates to v0.11.97 by @comfyui-wiki in #17368

New Contributors

• @curealb made their first contribution in #17221
• @DoronGenzelHass made their first contribution in #17224
• @Nynxz made their first contribution in #17220
• @yzeyu71 made their first contribution in #17359

Full Changelog: v0.42.0...v0.43.0

Fix some crashing issues a few users experienced when running the MOSS image template workflow.

What's Changed

• [Partner Nodes] feat(Krea): add Krea 3 Image node by @bigcat88 in #17067
• Update workflow templates to v0.11.85 by @comfyui-wiki in #17068
• Use FLUX 3 Image in node display labels by @MaanavD in #17075
• Better documentation about the comfy quant format. by @comfyanonymous in #17078
• Update workflow templates to v0.11.86 by @comfyui-wiki in #17083
• feat: Add MOSS-image support by @kijai in #17095
• [Partner Nodes] chore(Quiver): remove retired Arrow 1.x models by @bigcat88 in #17102
• Update comfy-kitchen version to 0.2.38 by @comfyanonymous in #17107
• Use fused situ with MOSS-image by @kijai in #17099
• Performance improvements for qwen 3.x (Part 2) by @rattus128 in #17007
• Update workflow templates to v0.11.87 by @comfyui-wiki in #17110
• Fix Qwen think-block templates in TextGenerate by @kijai in #17121
• Update workflow templates to v0.11.89 by @comfyui-wiki in #17138
• Add ARM Linux to execution and unit tests. by @comfyanonymous in #17139
• Refactor 3D model remeshing (CORE-560) by @kijai in #16805
• Fix crash issue with solid mode + empty mesh. by @comfyanonymous in #17140
• Bump comfyui-frontend-package to 1.55.18 by @comfy-pr-bot in #16859
• Fix webcam node no longer working. by @comfyanonymous in #17144
• add support for loading embeddings models onto Qwen3.5 by @silveroxides in #17159
• Support models with PDD baked in by @kijai in #17157
• Make NAF neighborhood attention work per LR cell by @kijai in #17155
• Fix text generation not properly cleaning itself up. by @comfyanonymous in #17169
• CPU device doesn't support fp8, nvfp4, mxfp8 or int8. by @comfyanonymous in #17170
• Update comfy-kitchen version to 0.2.39 by @comfyanonymous in #17171

New Contributors

• @MaanavD made their first contribution in #17075

Full Changelog: v0.41.0...v0.42.0

What's Changed

• Add ImageSDRToHDR node. by @comfyanonymous in #16990
• feat: Add a core route for the default workflow (CORE-340) by @alexisrolland in #16956
• Add ReferenceMultiLatent node. by @comfyanonymous in #17009
• Disable the LoRA stack nodes until the frontend supports DynamicGroup by @jaeone94 in #17050
• Improve ReferenceMultiLatent UX a bit. by @comfyanonymous in #17053
• Revert "Disable the LoRA stack nodes until the frontend supports DynamicGroup" by @comfyanonymous in #17054
• Support MiniMax-H3 LightVAE by @kijai in #17049
• Add nodes to get Latent Geometries by @rattus128 in #16093
• feat: Add Flux3-image model support (CORE-388) by @kijai in #17064
• Update workflow templates to v0.11.84 by @comfyui-wiki in #17065

Full Changelog: v0.40.0...v0.41.0

What's Changed

• [Server.py] Update CSP to allow blob: and data: by @S0yKaf in #15565
• Add comfy_attention and AttentionTensorContainer support to rest of models by @comfyanonymous in #16806
• Implement the asset export API locally (CORE-454) by @jtydhr88 in #16578
• Report 3D models from every 3D node as a standard 3d output item by @jtydhr88 in #16200
• feat(camera-info): CreateCameraInfo node + CameraInfoState widget input by @jtydhr88 in #14964
• feat(assets): extract video metadata into system_metadata on ingest and scan by @jtydhr88 in #15180
• Add a way to attach metadata to the websocket messages of a prompt. by @christian-byrne in #16763
• fix: preserve dynamic input validation and lazy execution by @jaeone94 in #16377
• refactor: keep DynamicGroup internal while stabilizing by @jaeone94 in #16811
• feat: add CameraAngle node with camera_info and shot prompt outputs by @jtydhr88 in #16251
• [Partner Nodes] feat(Google): add Nano Banana 2.1 model by @bigcat88 in #16822
• Update workflow templates to v0.11.77 by @comfyui-wiki in #16823
• Improve low VRAM performance of qwen 3.5 LLM by @rattus128 in #16820
• Make k/v contiguous before SDPA on gfx1151 when they don't fit in L2 by @tvukovic-amd in #16726
• Consolidate Linear forward API with prologue/epilogue fusion paths by @rattus128 in #16816
• Also attach the metadata to a prompt's preview image frames. by @christian-byrne in #16809
• Add governance enforcement for custom nodes by @guill in #16167
• comfy_attention configs in model files can now use kitchen sol attention. by @comfyanonymous in #16831
• Bump comfyui-frontend-package to 1.55.14 by @comfy-pr-bot in #16736
• [Partner Nodes] feat(Quiver): add V2 SVG nodes with per-model inputs and deprecate the originals by @bigcat88 in #16775
• [Partner Nodes] feat(Vidu): add nodes for Vidu Q4 model by @bigcat88 in #16849
• [Partner Nodes] feat(ClarityAI): add Crystal Upscaler node by @bigcat88 in #16817
• [Partner Nodes] feat(Kling): add Kling Virtual Try-On node by @bigcat88 in #16844
• Update workflow templates to v0.11.78 by @comfyui-wiki in #16851
• Change model code to use function wrappers for Linear calls by @rattus128 in #16861
• Fix issue with bypass loras and fused paths. by @comfyanonymous in #16862
• Use PNG compress level 4 for GLB textures for faster saving by @kijai in #16867
• Comfy Aimdo 0.5.6 + Fix qwen 2.1 rogues in low VRAM conditions by @rattus128 in #16870
• Fix Qwen3.5 MTP embedding offload by @rattus128 in #16877
• Loras can now replace the per layer attention config on a model. by @comfyanonymous in #16880
• Link related repositories in CodeRabbit reviews by @christian-byrne in #16769
• feat: add native LoRA stack loader nodes (CORE-186) by @jaeone94 in #16309
• Turn the assets system on by default; add --disable-assets by @synap5e in #16884
• [Partner Nodes] feat(HeyGen): add 2k resolution to HeyGen Video 1.0 nodes by @bigcat88 in #16876
• Add LatentOperationBlend and start/end percent to LatentApplyOperatio… by @comfyanonymous in #16923
• Disable the output assets scanning. by @comfyanonymous in #16926
• Enable MTP with images on Qwen3.5 by @kijai in #16922
• Decode MiniMax H3 single images as the first latent of a clip by @kijai in #16881
• fix(assets): date scanned assets by their file's mtime by @synap5e in #16810
• Enable compression on saved EXR. by @comfyanonymous in #16948
• Fix issues with loading half float files with pyav. by @comfyanonymous in #16949
• Keep linked repository findings focused by @christian-byrne in #16920
• Remove some outdated code. by @comfyanonymous in #16960
• Turn assets off unless --enable-assets, and scan the output folder again by @synap5e in #16929
• Turn the assets system on for the Windows standalone build by @synap5e in #16928
• Add a ARM64 nvidia portable. by @comfyanonymous in #16985
• Remove the cryptography package dependency. by @comfyanonymous in #16988
• Update embedded docs to v0.5.14 by @comfyui-wiki in #16981
• [Partner Nodes] feat(Higgsfield): Genjutsu Motion Transfer node by @bigcat88 in #16913
• feat(assets): previews for EXR assets, written by SaveImageAdvanced for outputs by @synap5e in #16898
• [Partner Nodes] feat(FishAudio): add Drama 3 model by @bigcat88 in #16999
• Update workflow templates to v0.11.82 by @comfyui-wiki in #17001
• Bump comfyui-frontend-package to 1.55.17 by @AustinMroz in #17004

New Contributors

• @S0yKaf made their first contribution in #15565

Full Changelog: v0.39.0...v0.40.0

What's Changed

• Support LynnReal light Minimax-H3 vae by @kijai in #16657
• Update embedded docs to v0.5.13 by @comfyui-wiki in #16618
• Use higher quality defaults for Save Video encoding. by @comfyanonymous in #16663
• feat: add DynamicGroup widget input by @jaeone94 in #16260
• fix(assets): don't let a model category that can't be listed abort the scan by @synap5e in #16658
• fix(assets): recover records when a drive comes back with hashing off by @synap5e in #16646
• Fix crash when selecting int8 or int4 cache in qwen image 2.1 by @comfyanonymous in #16667
• [Partner Nodes] feat(Anthropic): add Sonnet 5.5 model by @bigcat88 in #16647
• [Partner Nodes] feat(Ideogram): add nodes for Ideogram 4.5 model by @bigcat88 in #16689
• Fix minimax vae offload issue. by @comfyanonymous in #16698
• [Partner Nodes] feat(HeyGen): add nodes for HeyGen Video 1.0 model by @bigcat88 in #16695
• Update workflow templates to v0.11.73 by @comfyui-wiki in #16693
• Add 16-bit float support to Save EXR. by @comfyanonymous in #16706
• Bump comfyui-frontend-package to 1.53.10 by @comfy-pr-bot in #16456
• Release packed MiniMax H3 embedding temporaries before blocks by @Tokha233 in #16677
• Update default bit depth for 'exr' option to 16 bit float. by @comfyanonymous in #16715
• feat: add --offline and --disable-partner-nodes, deprecate --disable-api-nodes by @deepme987 in #16672
• [Partner Nodes] feat(BFL): add FLUX 3 Image node by @bigcat88 in #16716
• [Partner Nodes] feat(Grok): add grok-imagine-video-1.5-lite model by @bigcat88 in #16721
• Update workflow templates to v0.11.74 by @purzbeats in #16724
• fix(assets): batch prefix filters so scans work with many model folders by @synap5e in #16645
• Update comfy-kitchen version to 0.2.37 by @comfyanonymous in #16738
• [Partner Nodes] chore(Luma): deprecate Ray 2 nodes by @bigcat88 in #16741
• Update workflow templates to v0.11.76 by @comfyui-wiki in #16739
• Disable pinned memory automatically on integrated GPUs by @rattus128 in #16746
• feat(assets): report scan CPU and paused time, and classify scan failures by @synap5e in #16719
• fix(assets): don't take the database lock when assets are off; warn when it's held by @synap5e in #16742
• Echo the Create Bounding Boxes background as a UI preview by @jtydhr88 in #16636
• perf(assets): let the partial live-path index serve live-row-at-path lookups by @synap5e in #16659
• [Partner Nodes] feat(ElevenLabs): add Eleven v4 and v4 Turbo models by @bigcat88 in #16737
• [Partner Nodes] fix(OpenAI): remove transparent background for GPT Image 2 by @bigcat88 in #16744
• fix(assets): write the prune and offline marking in short batches so saves aren't locked out by @synap5e in #16696
• Add new attention and compiler stuff to AGENTS.md by @comfyanonymous in #16762
• Add comfy_attention and AttentionTensorContainer support to more models. by @comfyanonymous in #16766
• fix(assets): write scan inserts in short transactions so uploads and output saves aren't locked out + use WAL synchronous=NORMAL by @synap5e in #16748
• fix(assets): resume the background scan once the prompt queue is empty by @synap5e in #16765
• fix(assets): pause the startup scan during its folder walk and file stats by @synap5e in #16764
• fix(assets): copy uploads into place when the destination is on another volume by @synap5e in #16722

New Contributors

• @jaeone94 made their first contribution in #16260
• @Tokha233 made their first contribution in #16677

Full Changelog: v0.38.0...v0.39.0

What's Changed

• Improve qwen 2.1 KV cache location logic by @rattus128 in #16429
• Compile Qwen Image 2.1 transformer blocks by @rattus128 in #16430
• Allow model files to contain which attention should be used for each block. by @comfyanonymous in #16419
• JsonExtractString can now handle JSON with prefix and/or suffix. by @comfyanonymous in #16439
• Remove torchaudio dependency. by @comfyanonymous in #16457
• Cuda graphs + memory compiler on ace step 1.5 autoregressive model. by @comfyanonymous in #16461
• Blend H3 VAE tiles against composited neighbours by @kijai in #16436
• Fast disk detection to all model loaders. by @comfyanonymous in #16425
• [Partner Nodes] feat(Tencent): add Hunyuan Image 3.5 text-to-image and edit nodes by @comfyui-wiki in #16462
• Update workflow templates to v0.11.68 by @comfyui-wiki in #16466
• TextGenerate: Add system_prompt input and separate thinking output (CORE-460) by @kijai in #16442
• MiniMax-H3: Support Fun-Controlnet-Union 2.0 (CORE-462) by @kijai in #16471
• Fix MiniMax-H3 VAE rms_rope crash on offloaded qk_norm_scale by @chelsealong in #16485
• Port some optimizations to flux model family. by @comfyanonymous in #16488
• [Partner Nodes] feat(Quiver): add Arrow 2 models with reasoning effort widget by @bigcat88 in #16478
• [Partner Nodes] feat(Anthropic): add Claude Opus 5.5 to the Claude node by @bigcat88 in #16479
• [Partner Nodes] feat(Recraft): add Recraft V4.1 Flash to the V4 text-to-image node by @bigcat88 in #16501
• Update workflow templates to v0.11.69 by @comfyui-wiki in #16503
• Fix ImageUpscaleWithModel crashing on RGBA images by @chelsealong in #16500
• Take the SQLite write lock up front for asset scan and output-registration writes by @synap5e in #16480
• Move file reads out of the remaining asset write transactions by @synap5e in #16486
• feat: ming-image support by @kijai in #16482
• Fix saving/loading ming image model breaking detection. by @comfyanonymous in #16514
• Lower memory usage and .comfy_attention support for lumina family models. by @comfyanonymous in #16515
• [Partner Nodes] feat(OpenAI): add GPT-6 Sol and Luna models by @bigcat88 in #16520
• [Partner Nodes] feat(ByteDance): add Seedream 5.0 Flash by @bigcat88 in #16522
• fix(assets): harden the asset catalogue's failure paths by @synap5e in #16393
• Add missing RDNA2 arches to list. by @comfyanonymous in #16537
• feat(npu): support async weight offload streams by @Big2Wheel in #16057
• Small fix: skip unnecessary work. by @comfyanonymous in #16540
• Fix Qwen VL image preprocessing crash on 4-channel (RGBA) input by @chelsealong in #16548
• [Partner Nodes] feat(ByteDance): add Seedance 2.5 Draft mode by @bigcat88 in #16529
• Update workflow templates to v0.11.70 by @comfyui-wiki in #16557
• Add HDR LogC3 and ACEScct to Convert Color Space node. by @comfyanonymous in #16541
• Fix HDR output issue with negative values in Convert Image Color Space. by @comfyanonymous in #16565
• Update sheetsage2 abc creation from upstream yue code. by @comfyanonymous in #16569
• perf(assets): replace pathlib prefix checks in the asset scan by @synap5e in #16543
• Mention some ComfyUI optimization details in README. by @comfyanonymous in #16576
• Support tiny VAE for Qwen-Image 2.1 by @kijai in #16552
• feat: Support ID-V2V Wan 2.1/VACE based model (CORE-426) by @kijai in #15139
• Fix potential regression with previous PR. by @comfyanonymous in #16596
• [Partner Nodes] chore(Sora): remove deprecated Sora nodes by @bigcat88 in #16609
• Clamp Qwen Image 2.1 fp16 activations in place by @dcbert in #16608
• Add comfy_attention and AttentionTensorContainer to a few models. by @comfyanonymous in #16595
• Speedup Yue2 AR by @rattus128 in #16626
• Searching "kitchen" now returns the attention node. by @comfyanonymous in #16632
• Support Qwen-Image 2.1 union fun controlnet by @kijai in #16519
• Update comfy-kitchen package version to 0.2.36 by @comfyanonymous in #16635
• Optimize Seedvr2 by @kijai in #16530
• Support w6a8 quantization format by @kijai in #16483
• Fixes for last PR. by @comfyanonymous in #16637
• fix(db): wait briefly for the database lock at startup by @synap5e in #16602
• fix(assets): Start without the assets packages; --enable-assets reports what is missing and runs with assets disabled by @synap5e in #16580
• perf(assets): keep the UI responsive during asset scans and rescan output by listing folders by @synap5e in #16546
• Speedup generation on qwen3.5/3.8 with long contexts. by @comfyanonymous in #16638
• Avoid unnecessary write. by @comfyanonymous in #16639
• Update codeowners. by @comfyanonymous in #16655

New Contributors

• @Big2Wheel made their first contribution in #16057
• @dcbert made their first contribution in #16608

Full Changelog: v0.37.0...v0.38.0

What's Changed

• Aimdo 0.5.5 + Auto-detect and enable --fast-disk when the disk is fast (CORE-440) by @rattus128 in #16333
• [Partner Nodes] feat(OpenAI): add transparent background support for GPT Image 2 by @bigcat88 in #16366
• Add CFG control to YuE2 Generate Music node. by @comfyanonymous in #16373
• Always put text encoder on GPU when dynamic vram on. by @comfyanonymous in #16374
• [Partner Nodes] fix(Tripo): refuse a P2 run whose linked GLB or FBX output would be empty by @bigcat88 in #16369
• feat: Support MoGe 3 (CORE-443) by @kijai in #16381
• Update comfy-kitchen version to 0.2.35 by @comfyanonymous in #16385
• chores: Update default width and height in EmptyLatentImage by @alexisrolland in #16384
• chores: Update nodes names and categories by @alexisrolland in #16274
• Qwen3/3.5/3.8 cudagraphs and w4a8 gemv support (CORE-390) by @kijai in #15623
• Fix issue with qwen speedup PR. by @comfyanonymous in #16389
• Lower the pos embed precision of SheetSage2 to match upstream. by @comfyanonymous in #16395
• [Partner Nodes] feat(client): send Idempotency-Key on partner-proxy calls and collect replays by @bigcat88 in #16220
• Bump comfyui-frontend-package to 1.53.6 by @comfy-pr-bot in #16386
• Fix ace step VAE decode crashing on non bf16 GPUs. by @comfyanonymous in #16405
• feat: Qwen-image 2.1 support (CORE-423) by @kijai in #16400
• Update embedded docs to v0.5.12 by @comfyui-wiki in #16396
• Lower wan peak vram when using comfy kitchen attention. by @comfyanonymous in #16418
• Update workflow templates to v0.11.65 by @comfyui-wiki in #16423
• Add a way to disable fast disk for debugging. by @comfyanonymous in #16424
• Update workflow templates to v0.11.66 by @comfyui-wiki in #16426
• [Partner Nodes] feat(Meshy): add Meshy 7.1 support by @bigcat88 in #16412
• Fix MiniMax Music 3 producing noise with CUDA graphs enabled. by @RaynoldVanHeyningen in #16428

New Contributors

• @RaynoldVanHeyningen made their first contribution in #16428

Full Changelog: v0.36.0...v0.37.0