v0.3.49

What's Changed

• API for Recently Used Items by @Myestery in #8792
• Update template to 0.1.47 by @comfyui-wiki in #9153
• Make sure all the conds are on the right device. by @comfyanonymous in #9151
• Make sure context is on the right device. by @comfyanonymous in #9154
• Cleanup. by @comfyanonymous in #9160
• Lower cond vram use by casting at the same time as device transfer. by @comfyanonymous in #9159
• Fix broken controlnet from last PR. by @comfyanonymous in #9167
• Various fixes for broken things from earlier PR. by @comfyanonymous in #9168
• Add some warnings and prevent crash when cond devices don't match. by @comfyanonymous in #9169
• Fix another issue with the PR. by @comfyanonymous in #9170
• Initial support for qwen image model. by @comfyanonymous in #9179
• Add Veo3 video generation node with audio support by @christian-byrne in #9110
• Update template to 0.1.48 by @comfyui-wiki in #9182
• Fix default shift and any latent size for qwen image model. by @comfyanonymous in #9186
• Update template to 0.1.51 by @comfyui-wiki in #9187

New Contributors

• @Myestery made their first contribution in #8792

Full Changelog: v0.3.48...v0.3.49

v0.3.48

What's Changed

• ComfyAPI Core v0.0.2 by @guill in #8962
• Update template to 0.1.44 by @comfyui-wiki in #9104
• Small wan camera memory optimization. by @comfyanonymous in #9111
• Replace manual x * sigmoid(x) with torch silu in VAE nonlinearity by @chaObserv in #9057
• Add support for partial execution in backend by @guill in #9123
• More accurate explanation of release process. by @comfyanonymous in #9126
• V3 Node Schema Definition - initial by @Kosinkadink in #8656
• Fix WanFirstLastFrameToVideo node when no clip vision. by @comfyanonymous in #9134
• update template to 0.1.45 by @comfyui-wiki in #9135
• Tiny wan vae optimizations. by @comfyanonymous in #9136

Full Changelog: v0.3.47...v0.3.48

v0.3.47

What's Changed

• Remove unecessary clones in the wan2.2 VAE. by @comfyanonymous in #9083
• Extra reserved vram on large cards on windows. by @comfyanonymous in #9093
• Make wan2.2 5B i2v take a lot less memory. by @comfyanonymous in #9102

Full Changelog: v0.3.46...v0.3.47

v0.3.46

What's Changed

• Replace torchaudio.load with pyav. by @comfyanonymous in #8989
• Add hidream e1.1 example to readme. by @comfyanonymous in #8990
• This is not needed. by @comfyanonymous in #8991
• Try to fix line endings workflow. by @comfyanonymous in #9001
• Add xpu version for async offload and some other things. by @simonlui in #9004
• Add default device argument. by @comfyanonymous in #9023
• Fix xpu function not implemented. by @comfyanonymous in #9026
• Fix xpu function not implemented p2. by @comfyanonymous in #9027
• Enable pytorch attention by default for gfx1201 on torch 2.8 by @comfyanonymous in #9029
• Only enable cuda malloc on cuda torch. by @comfyanonymous in #9031
• Separate denoised and noise estimation in Euler CFG++ by @chaObserv in #9008
• [Training Node] algo support, grad acc, optional grad ckpt by @KohakuBlueleaf in #9015
• Support Iluvatar CoreX by @honglyua-il in #8585
• Added parameter required_frontend_version in the /system_stats API response by @shivansh-gupta4 in #8875
• Print xpu device name. by @comfyanonymous in #9035
• Add WAN ATI support by @ecfairle in #8874
• Remove useless code. by @comfyanonymous in #9041
• Update template to 0.1.40 by @comfyui-wiki in #9048
• Fix issue with line endings github workflow. by @comfyanonymous in #9053
• Update negative prompt for Moonvalley nodes by @Thor-ATX in #9038
• Add map_function to get_history. by @comfyanonymous in #9056
• Remove useless code. by @comfyanonymous in #9059
• Fix link from docs (due to document updates) by @comfyui-wiki in #9060
• Small readme update. by @comfyanonymous in #9071
• Add Moonvalley Marey V2V node with updated input validation by @christian-byrne in #9069
• Update template to 0.1.41 by @comfyui-wiki in #9079
• Wan 2.2 support. by @comfyanonymous in #9080
• Add wan2.2 to readme. by @comfyanonymous in #9081

New Contributors

• @honglyua-il made their first contribution in #8585
• @shivansh-gupta4 made their first contribution in #8875
• @ecfairle made their first contribution in #8874
• @Thor-ATX made their first contribution in #9038

Full Changelog: v0.3.45...v0.3.46

v0.3.45

What's Changed

• Un-hardcode Chroma patch_size by @josephrocca in #8840
• Add reconstructed SA-Solver sampler by @chaObserv in #8834
• Support Multi Image-Caption dataset in lora training node by @KohakuBlueleaf in #8819
• Remove unmaintained notebook. by @comfyanonymous in #8845
• Better training loop implementation by @KohakuBlueleaf in #8820
• Add warning to catch torch import mistakes. by @comfyanonymous in #8852
• Support for async node functions by @guill in #8830
• Fix python3.9 by @comfyanonymous in #8858
• [fix] increase Kling API polling timeout to prevent user timeouts by @christian-byrne in #8860
• Remove windows line endings. by @comfyanonymous in #8866
• Execute async node earlier by @JettHu in #8865
• Fix fresca's input and output by @chaObserv in #8871
• Add model detection error hint for lora. by @comfyanonymous in #8880
• Remove tokens from history items by @christian-byrne in #8889
• Refactor previous pr. by @comfyanonymous in #8893
• Remove windows line endings. by @comfyanonymous in #8902
• Upate template to 0.1.36 by @comfyui-wiki in #8904
• Allow the prompt request to specify the prompt ID. by @FeepingCreature in #8189
• Fix wrong reference bug by @yoland68 in #8910
• Update nodes_gemini.py by @yoland68 in #8912
• Add action to detect windows line endings. by @comfyanonymous in #8917
• Fix MaskComposite error when destination has 2 dimensions by @bman654 in #8915
• Add attempt to work around the safetensors mmap issue. by @comfyanonymous in #8928
• LTXV VAE decoder: switch default padding mode by @harelc in #8930
• Silence clip tokenizer warning. by @comfyanonymous in #8934
• Experimental CFGNorm node. by @comfyanonymous in #8942
• Forgot this by @comfyanonymous in #8957
• Add nested style of dual cfg to DualCFGGuider node. by @comfyanonymous in #8965
• Update template to 0.1.37 by @comfyui-wiki in #8967
• Add SamplingPercentToSigma node by @chaObserv in #8963
• Document what the fast_fp16_accumulation is in the portable. by @comfyanonymous in #8973
• Update template to 0.1.39 by @comfyui-wiki in #8981

New Contributors

• @josephrocca made their first contribution in #8840
• @bman654 made their first contribution in #8915

Full Changelog: v0.3.44...v0.3.45

v0.3.44

What's Changed

• “--whitelist-custom-nodes” args for comfy core to go with “--disable-all-custom-nodes” for development purposes by @lgldlk in #8592
• Skip custom node logic completely if disabled and no whitelisted nodes. by @comfyanonymous in #8719
• Reorder the schedulers so simple is the default one. by @comfyanonymous in #8722
• Perpneg guider with updated pre and post-cfg by @chaObserv in #8698
• Make stable release workflow publish drafts. by @comfyanonymous in #8723
• [System Pop Up] Add GitHub Action for release webhook notifications by @bmcomfy in #8652
• Fix contiguous issue with pytorch nightly. by @comfyanonymous in #8729
• Dual cfg node optimizations when cfg is 1.0 by @comfyanonymous in #8747
• Update template to 0.1.31 by @comfyui-wiki in #8741
• Add TCFG node by @chaObserv in #8730
• Migrate ER-SDE from VE to VP algorithm and add its sampler node by @chaObserv in #8744
• PerpNeg Guider optimizations. by @comfyanonymous in #8753
• support upload 3d model to custom subfolder by @jtydhr88 in #8597
• Fallback to regular op when fp8 op throws exception. by @comfyanonymous in #8761
• nodes_lt: fixes to latent conditioning at index > 0 by @harelc in #8769
• Fix ImageColorToMask not returning right mask values. by @comfyanonymous in #8771
• Alternative Skip Layer Guidance (SLG) node implementation by @city96 in #8759
• Use torch cu129 for nvidia pytorch nightly. by @comfyanonymous in #8786
• Update template to 0.1.32 by @comfyui-wiki in #8782
• Add ImageRotate and ImageFlip nodes. by @comfyanonymous in #8789
• Add the denoising step to several samplers by @chaObserv in #8780
• Add that ckpt files are loaded safely to README. by @comfyanonymous in #8791
• Add warning when loading file unsafely. by @comfyanonymous in #8800
• Remove some useless code. by @comfyanonymous in #8812
• Remove dependency on deprecated torchaudio.save function by @comfyanonymous in #8815
• Fix warning in audio save nodes. by @comfyanonymous in #8818
• Update template and node docs package version by @comfyui-wiki in #8825
• Update template to 0.1.34 by @christian-byrne in #8829
• Update template to 0.1.35 by @comfyui-wiki in #8831
• Add Moonvalley nodes by @christian-byrne in #8832

New Contributors

• @lgldlk made their first contribution in #8592
• @bmcomfy made their first contribution in #8652
• @harelc made their first contribution in #8769

Full Changelog: v0.3.43...v0.3.44

v0.3.43

What's Changed

• Add Flux Kontext and Omnigen 2 models to readme. by @comfyanonymous in #8682
• Don't add tiny bit of random noise when VAE encoding. by @comfyanonymous in #8705
• Fix memory estimation bug with kontext. by @comfyanonymous in #8709

Full Changelog: v0.3.42...v0.3.43

v0.3.42

What's Changed

Major new things in this release are the Flux Kontext and Omnigen2 image editing models.

• Add Cosmos Predict2 to README. by @comfyanonymous in #8562
• Update template to 0.1.29 by @comfyui-wiki in #8563
• Update frontend to 1.22.2 by @webfiltered in #8567
• Show a better error when the workflow OOMs. by @comfyanonymous in #8574
• Add new fields to the config types by @coderfromthenorth93 in #8507
• Fix pytorch warning. by @comfyanonymous in #8593
• Make flux controlnet work with sd3 text enc. by @comfyanonymous in #8599
• Small flux optimization. by @comfyanonymous in #8611
• Improve the log time function for 10 minute + renders by @tonynoce in #6207
• Fix exception when using a noise mask with cosmos predict2. by @comfyanonymous in #8621
• Allow padding in ImageStitch node to be white. by @comfyanonymous in #8631
• Resize and pad image node. by @comfyanonymous in #8636
• Indicate that directml is not recommended in the README. by @comfyanonymous in #8644
• Cosmos predict2 model merging nodes. by @comfyanonymous in #8647
• Singlestep DPM++ SDE for RF by @chaObserv in #8627
• unet -> diffusion model by @comfyanonymous in #8659
• Omnigen2 model implementation. by @comfyanonymous in #8669
• Bump minimum transformers version. by @comfyanonymous in #8671
• Disable omnigen2 fp16 on older pytorch versions. by @comfyanonymous in #8672
• Support Flux Kontext Dev model. by @comfyanonymous in #8679
• Update frontend to 1.23.3 by @webfiltered in #8678
• Update requirements.txt by @comfyanonymous in #8680
• Update frontend to 1.23.4 by @webfiltered in #8681

New Contributors

• @coderfromthenorth93 made their first contribution in #8507
• @tonynoce made their first contribution in #6207

Full Changelog: v0.3.41...v0.3.42

v0.3.41

What's Changed

• Let chroma TE work on regular flux. by @comfyanonymous in #8429
• chore: add badge and link to social network Twitter by @Olexandr88 in #8427
• Alternate fix for #8435 by @comfyanonymous in #8442
• Cleaner torch version comparisons. by @comfyanonymous in #8453
• Enable fp8 ops by default on gfx1201 by @comfyanonymous in #8464
• Enable AMD fp8 and pytorch attention on some GPUs. by @comfyanonymous in #8474
• Fix WebcamCapture IS_CHANGED signature by @kmewhort in #8413
• Apple most likely is never fixing the fp16 attention bug. by @comfyanonymous in #8485
• Update template to 0.1.27 by @comfyui-wiki in #8487
• Fix black images on some flux models in fp16. by @comfyanonymous in #8495
• Fix mistake in last commit. by @comfyanonymous in #8496
• auto register web folder from pyproject by @jtydhr88 in #8478
• Add support for sqlite database by @pythongosssss in #8444
• Revert "auto register web folder from pyproject (#8478)" by @comfyanonymous in #8497
• Allow specifying any frontend semver suffixes by @webfiltered in #8498
• Revert "Update template to 0.1.27" by @webfiltered in #8502
• Add a warning for old python versions. by @comfyanonymous in #8504
• Auto register web folder by @jtydhr88 in #8505
• debug -> warning by @comfyanonymous in #8506
• Add '@prerelease' to use latest test frontend by @webfiltered in #8501
• Fix Kontext aspect ratio only works as widget by @christian-byrne in #8508
• Support cosmos predict2 text to image 2B and 14B models. by @comfyanonymous in #8517
• Uncap cosmos predict2 res and fix mem estimation. by @comfyanonymous in #8518
• Add cosmos_rflow option to ModelSamplingContinuousEDM node. by @comfyanonymous in #8523
• LoRA Trainer: LoRA training node in weight adapter scheme by @KohakuBlueleaf in #8446
• Bump embedded docs to 0.2.2 by @comfyui-wiki in #8512
• Bump template to 0.1.28 by @comfyui-wiki in #8510
• allow extra settings from pyproject.toml by @jtydhr88 in #8526
• Small cosmos attention code refactor. by @comfyanonymous in #8530
• Generalize SEEDS samplers for RF by @chaObserv in #8529
• Support Cosmos predict2 image to video models. by @comfyanonymous in #8535
• SaveLora now saves in the same filename format as all the other nodes. by @comfyanonymous in #8538
• Add correct eps to ltxv rmsnorm. by @comfyanonymous in #8542
• Allow setting min_length with tokenizer_data. by @comfyanonymous in #8547
• Multistep DPM++ SDE samplers for RF by @chaObserv in #8541

New Contributors

• @Olexandr88 made their first contribution in #8427
• @kmewhort made their first contribution in #8413

Full Changelog: v0.3.40...v0.3.41

v0.3.40

What's Changed

• Cleanup BFL API nodes for Kontext [pro] and [max] by @Kosinkadink in #8337
• Add node for regex replace(sub) operation by @JettHu in #8340
• Put ROCm version in tuple to make it easier to enable stuff based on it. by @comfyanonymous in #8348
• Remove huchenlei from CODEOWNERS by @huchenlei in #8350
• fix: custom comfy-api-base works with subpath by @BennyKok in #8332
• use fused multiply-add pointwise ops in chroma by @drhead in #8279
• Bump template to 0.1.23 by @comfyui-wiki in #8353
• Make it easier to pass lists of tensors to models. by @comfyanonymous in #8358
• Update frontend to 1.21 by @webfiltered in #8366
• [feat] Add ImageStitch node for concatenating images by @christian-byrne in #8369
• Add Help Menu in NodeLibrarySidebarTab by @benceruleanlu in #8179
• Make the casting in lists the same as regular inputs. by @comfyanonymous in #8373
• Bump template to 0.1.25 by @comfyui-wiki in #8372
• [BugFix] Update frontend to 1.21.4 by @webfiltered in #8377
• [feat] Add custom node testing requirement to issue templates by @christian-byrne in #8374
• [BugFix] Update frontend to 1.21.5 by @webfiltered in #8382
• [BugFix] Update frontend to 1.21.6 by @webfiltered in #8383
• Update fix for potential XSS on /view by @jessegonyou in #8384
• Style fix. by @comfyanonymous in #8390
• [feat] Add GetImageSize node by @christian-byrne in #8386
• Add api nodes to readme. by @comfyanonymous in #8402
• add support to read pyproject.toml from custom node by @jtydhr88 in #8357
• Update frontend to 1.21.7 by @webfiltered in #8410
• [Fix] Replace call to deprecated pillow API Image.ANTIALIAS by @soudfl in #8415
• Add batch to GetImageSize node. by @comfyanonymous in #8419
• [refactor] remove version prefixes from Ideogram node categories by @christian-byrne in #8418

New Contributors

• @BennyKok made their first contribution in #8332
• @jessegonyou made their first contribution in #8384
• @soudfl made their first contribution in #8415

Full Changelog: v0.3.39...v0.3.40