8 Releases published by 1 person

• v0.3.19

  published Mar 4, 2025

• v0.3.20

  published Mar 5, 2025

• v0.3.21

  published Mar 5, 2025

• v0.3.22

  published Mar 5, 2025

• v0.3.23

  published Mar 6, 2025

• v0.3.24

  published Mar 6, 2025

• v0.3.25

  published Mar 9, 2025

• v0.3.26

  published Mar 9, 2025

11 Pull requests merged by 7 people

• ltxv: relax frame_idx divisibility for single frames.

  #7146 merged Mar 10, 2025

• [3d] remove unused params

  #6931 merged Mar 9, 2025

• Weight Hooks Switching Optimization

  #7067 merged Mar 9, 2025

• Update frontend to 1.11.8

  #7119 merged Mar 8, 2025

• Typo in node_typing.py

  #7092 merged Mar 6, 2025

• fixed: Incorrect guide message for missing frontend.

  #7105 merged Mar 6, 2025

• Add type hint for FileLocator

  #6968 merged Mar 5, 2025

• Better argument handling of front-end-root

  #7043 merged Mar 5, 2025

• [NodeDef] Explicitly add control_after_generate to seed/noise_seed

  #7059 merged Mar 5, 2025

• fix: ltxv crop guides works with 0 keyframes

  #7085 merged Mar 5, 2025

• Improve: Provide a better guide message for the missing frontend message.

  #7079 merged Mar 5, 2025

9 Pull requests opened by 9 people

• MultiGPU Work Units For Accelerated Sampling

  #7063 opened Mar 4, 2025

• Add TE support for Lumina 2 LoRA

  #7075 opened Mar 4, 2025

• LTXVVideoInterpolator

  #7091 opened Mar 6, 2025

• [Draft] Native LotusD Implementation

  #7125 opened Mar 7, 2025

• Fix LoadImageOutput node

  #7143 opened Mar 9, 2025

• make conds use non-blocking transfers

  #7152 opened Mar 9, 2025

• General way of receiving image for multiple/unknown json with SaveImageWebsocket

  #7160 opened Mar 10, 2025

• Prevent custom nodes from accidentally overwriting global modules.

  #7167 opened Mar 10, 2025

• Only check frontend package if using default frontend

  #7179 opened Mar 10, 2025

46 Issues closed by 22 people

• Group Node Export/Import Failure

  #7174 closed Mar 10, 2025

• 有没有大佬知道怎么解决

  #7153 closed Mar 10, 2025

• After update, comfyui manager is disappeared

  #7165 closed Mar 10, 2025

• Commit broke loading of custom nodes - Resolved. Update to b779349b55e79aff81a98b752f5cb486c71812db

  #7162 closed Mar 10, 2025

• 5070ti

  #7159 closed Mar 10, 2025

• new release big workflow slow and not work

  #7060 closed Mar 10, 2025

• "Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same" [Since v0.3.15]

  #7135 closed Mar 9, 2025

• BUG

  #7144 closed Mar 9, 2025

• Why is this? Has any brother encountered such a problem

  #7130 closed Mar 9, 2025

• How to make comfyui load a custom default workflow (docker)?

  #7107 closed Mar 9, 2025

• 【求助】mat1 and mat2 shapes cannot be multiplied (77x768 and 4096x5120)

  #7114 closed Mar 9, 2025

• Newest frontend update (partially) broke previews

  #7134 closed Mar 8, 2025

• Disable SageAttention so I can use SD 1.5?

  #7140 closed Mar 8, 2025

• Option to unlock full live preview when generate image like older version

  #7139 closed Mar 8, 2025

• load workflow from image broke ui;version0.3.24

  #7132 closed Mar 8, 2025

• The load image node that comes with ComfyUI is missing a line called "choose file to upload"

  #7138 closed Mar 8, 2025

• Blinking effect from Wan video output due to Vae Decode (Tiled)

  #7116 closed Mar 8, 2025

• RuntimeError: Input type (torch.FloatTensor) and weight type (XPUFloatType) should be the same or input should be a MKLDNN tensor and weight is a dense tensor

  #7089 closed Mar 7, 2025

• 求求

  #7122 closed Mar 7, 2025

• ComfyUI stopped working after update to 3.24

  #7120 closed Mar 7, 2025

• Loading image node does not show preview

  #7123 closed Mar 7, 2025

• Generation history

  #7121 closed Mar 7, 2025

• 更新最新版本报错has no attribute 'LTXVAddGuide'

  #7113 closed Mar 7, 2025

• k采样器错误

  #6949 closed Mar 7, 2025

• Adding a Stable version

  #7088 closed Mar 7, 2025

• Image generation fundamentally changed after updating ComfyUI

  #7110 closed Mar 7, 2025

• Exception during processing !!! expected sequence of length 69 at dim 1 (got 77)

  #7102 closed Mar 6, 2025

• How to solve the problem of Kernel size can't be greater than actual input size?

  #7095 closed Mar 6, 2025

• WanVideoSampler///Empty image embeds must be provided for T2V (Text to Video)

  #7100 closed Mar 6, 2025

• Changable Fonts

  #3085 closed Mar 6, 2025

• If the clip is from a checkpoint loader node your checkpoint does not contain a valid clip or text encoder model.

  #7080 closed Mar 6, 2025

• Add workflow auto-save switch .

  #7081 closed Mar 6, 2025

• 反推提示词：Cannot connect to host huggingface.co:443 ssl:default [Connection reset by peer]

  #7094 closed Mar 6, 2025

• Desktop ComfyUI (Windows) executable won't accept parameters on it's commandline

  #7087 closed Mar 5, 2025

• load image No components choose image

  #7078 closed Mar 5, 2025

• Google Colab ComfyUI run with cloudflared

  #6941 closed Mar 5, 2025

• value slider multiline tooltip covers the value itself, needs to be repositioned 20px lower

  #7062 closed Mar 4, 2025

• UserWarning: Synchronous compression of large response bodies (1389192 bytes) might block the async event loop

  #7066 closed Mar 4, 2025

• facing problem

  #7069 closed Mar 4, 2025

• ControlNetFlux.forward() missing 1 required positional argument: 'y'

  #7070 closed Mar 4, 2025

• Wan Image2Video - 'VAE' object has no attribute 'vae_dtype'

  #7053 closed Mar 4, 2025

• Nvidia 3090 CUDA error

  #7064 closed Mar 4, 2025

• invalid style model

  #7050 closed Mar 4, 2025

• how to upload a video to cpmfyui via api

  #4352 closed Mar 4, 2025

• Unable to load nvfuser_codegen.dll

  #4391 closed Mar 4, 2025

• Lost manager when install custom nodes

  #7003 closed Mar 4, 2025

56 Issues opened by 55 people

• The size of tensor a (49) must match the size of tensor b (16) at non-singleton dimension 1

  #7177 opened Mar 10, 2025

• RAM is loaded too much(>33gb) while using LoRA with flux dev. When Lora is not used - RAM loaded only for 2gb.

  #7176 opened Mar 10, 2025

• 5070tI 錯誤

  #7175 opened Mar 10, 2025

• Save finished queue like pair png + json workflow

  #7173 opened Mar 10, 2025

• Problem with generation image with simple workflow

  #7172 opened Mar 10, 2025

• MPS backend out of memory (MPS allocated: 16.91 GB, other allocations: 2.77 MB, max allowed: 18.13 GB). Tried to allocate 1.28 GB on private pool. Use PYTORCH_MPS_HIGH_WATERMARK_RATIO=0.0 to

  #7171 opened Mar 10, 2025

• Problem with Comfy-Manager and rgthree nodes after last update

  #7169 opened Mar 10, 2025

• RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory

  #7168 opened Mar 10, 2025

• Error loading extension /extensions

  #7166 opened Mar 10, 2025

• i was using flux and this happen what shod i do

  #7164 opened Mar 10, 2025

• Sometime crash in wan2.1 only change the prompt

  #7163 opened Mar 10, 2025

• IPAdapterFaceID加载器

  #7161 opened Mar 10, 2025

• 更新后执行工作流没有任何反应，

  #7158 opened Mar 10, 2025

• chrash after change sdxl model on rtx 3050

  #7156 opened Mar 10, 2025

• After installing triton and sageattention, restarting comfyui resulted in errors with custom nodes. Please help from all the experts, thank you.

  #7154 opened Mar 10, 2025

• Failed to validate prompt

  #7151 opened Mar 9, 2025

• ComfyUI worked good a 2 month later, but after reinstall they crashed every time

  #7150 opened Mar 9, 2025

• Wan2.1 image to workflow video gives me this error

  #7149 opened Mar 9, 2025

• ComfyUI portable tablets version. How to "D:\a\ComfyUI\python_embeded\python.exe" Change to actual path

  #7148 opened Mar 9, 2025

• SingleStreamBlock.forward() got an unexpected keyword argument 'modulation_dims_img'

  #7147 opened Mar 9, 2025

• No such file or directory: 'I:\\AI\\ComfyUI\\output\\ComfyUI_02638_.png [output]'

  #7142 opened Mar 9, 2025

• FLUX Training lora error

  #7141 opened Mar 9, 2025

• FETCH ComfyRegistry Data & database is locked reading lora bug !!!

  #7137 opened Mar 8, 2025

• help plz

  #7136 opened Mar 8, 2025

• Generated Images Gradually Deviate from Prompt or LoRA Constraints in Official Flux Text-to-Image Workflow

  #7133 opened Mar 8, 2025

• Custom naming for multiple comfyUI instances

  #7131 opened Mar 8, 2025

• macOS App - Adjust paths for saving

  #7128 opened Mar 8, 2025

• UserWarning: NVIDIA GeForce RTX 5070 Ti with CUDA capability sm_120 is not compatible with the current PyTorch installation. The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_61 sm_70 sm_75 sm_80 sm_86 sm_90.

  #7127 opened Mar 8, 2025

• CUDA

  #7126 opened Mar 8, 2025

• Bad File Descriptor Errors During Model Loading

  #7124 opened Mar 7, 2025

• AdvancedControlBase.get_control_inject() takes 5 positional arguments but 6 were given

  #7118 opened Mar 7, 2025

• I hope add --path-prefix args, like tensorboard.

  #7117 opened Mar 7, 2025

• CLIPTextEncode

  #7115 opened Mar 7, 2025

• subprocess.CalledProcessError | KSampler report : Command ~MSVC\\14.43.34808\\bin\\Hostx64\\x64\\cl.exe

  #7112 opened Mar 7, 2025

• Specify content type in executed output messages

  #7111 opened Mar 6, 2025

• Using an embedding: with different architecture now cause an error.

  #7109 opened Mar 6, 2025

• Screen freezes when running ksampler

  #7108 opened Mar 6, 2025

• cannot import name 'AttrsDescriptor' when using TorchCompileModel node

  #7106 opened Mar 6, 2025

• Pause, Press anykey to continue

  #7104 opened Mar 6, 2025

• Failed to install python packages

  #7103 opened Mar 6, 2025

• Interface/instance kind of node representation

  #7099 opened Mar 6, 2025

• 远程主机强迫关闭了一个现有的连接是什么原因？

  #7097 opened Mar 6, 2025

• Getting 'list index out of range' Error in ComfyUI on Mac Studio - Any Tips?"

  #7096 opened Mar 6, 2025

• Add ACE plus unified model native support

  #7093 opened Mar 6, 2025

• Offering help with Unit Tests

  #7086 opened Mar 5, 2025

• 报错

  #7084 opened Mar 5, 2025

• Load workflows

  #7083 opened Mar 5, 2025

• For the problem that images with the same file name in the loading image window will not be refreshed?

  #7082 opened Mar 5, 2025

• cogview4

  #7077 opened Mar 5, 2025

• Problem

  #7073 opened Mar 4, 2025

• ComfyUI Interface frozen.

  #7072 opened Mar 4, 2025

• The IP adapter model cannot be released from VRAM

  #7071 opened Mar 4, 2025

• RX 7900 XTX (maybe AMD in general) render time build-up issue

  #7068 opened Mar 4, 2025

• Florence2Model get error report

  #7061 opened Mar 4, 2025

• SamplerCustomAdvanced error, Node ID 13

  #7058 opened Mar 4, 2025

• comfyui 3dpack run unique3d work flow，warning this

  #7057 opened Mar 3, 2025

57 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Add disable_mmap arg in method load_torch_file

  #6981 commented on Mar 7, 2025 • 8 new comments

• Feat: Add basic LoRA training support

  #7032 commented on Mar 5, 2025 • 3 new comments

• Dockerfile and automatic generation of images via GitHub Actions

  #1469 commented on Mar 10, 2025 • 2 new comments

• torch.OutOfMemoryError: Allocation on device

  #6856 commented on Mar 3, 2025 • 0 new comments

• ComfyUI Startup Log on macOS Sequoia 15.1.1 with MPS and PyTorch (BFloat16 Unsupported)

  #5829 commented on Mar 7, 2025 • 0 new comments

• 报错了,一直没弄明白什么原因,麻烦看下谢谢

  #5751 commented on Mar 7, 2025 • 0 new comments

• 我的ControlNet应用出现了一些问题

  #6956 commented on Mar 7, 2025 • 0 new comments

• Model reloads for every API request, adding additional 15sec to each inference request. Issue with indexing.

  #6898 commented on Mar 7, 2025 • 0 new comments

• Speed up the first run.

  #1992 commented on Mar 7, 2025 • 0 new comments

• flux torch cudagraphs compilation broken again on 2.6.0

  #6866 commented on Mar 7, 2025 • 0 new comments

• Pause and Re-arrange Queue

  #6710 commented on Mar 8, 2025 • 0 new comments

• Feature request: Pause Queue

  #1032 commented on Mar 9, 2025 • 0 new comments

• Torch Model Compile does not work with lora

  #5375 commented on Mar 9, 2025 • 0 new comments

• [Announcement] The frontend will no longer be shipped in the main ComfyUI repo, it will be a separate pip package instead.

  #7025 commented on Mar 9, 2025 • 0 new comments

• TypeError: forward_timestep_embed() got an unexpected keyword argument 'time_context'

  #2037 commented on Mar 9, 2025 • 0 new comments

• the group node display is problematic.

  #6967 commented on Mar 9, 2025 • 0 new comments

• How to extend LoadImage class? or any node class?

  #1296 commented on Mar 10, 2025 • 0 new comments

• About git action execute error

  #6801 commented on Mar 10, 2025 • 0 new comments

• How to composite noise per region in regional sampling?

  #6984 commented on Mar 10, 2025 • 0 new comments

• I get an "AttributeError: 'tuple' object has no attribute 'clone'" when every I attempt to use basic "merge model" nodes

  #6495 commented on Mar 10, 2025 • 0 new comments

• XPU out of memory Error after LoadImage and KSampler

  #7004 commented on Mar 10, 2025 • 0 new comments

• BatchPromptSchedule

  #7038 commented on Mar 10, 2025 • 0 new comments

• Extremely slow image generation with Flux model on MacOS Apple M4 Pro

  #6969 commented on Mar 10, 2025 • 0 new comments

• Run with CPU automatically when GPU is unavailable

  #6541 commented on Mar 10, 2025 • 0 new comments

• High lora vram usage after update

  #4343 commented on Mar 10, 2025 • 0 new comments

• Wan model is not working in MacOs if scheduler is `uni_pc`

  #7027 commented on Mar 10, 2025 • 0 new comments

• Lora OneTrainer for Hunyuan video compatible with ComfyUI?

  #6531 commented on Mar 10, 2025 • 0 new comments

• Add OpenVINO backend for torch.compile node

  #6638 commented on Mar 7, 2025 • 0 new comments

• add docker config file

  #6993 commented on Mar 10, 2025 • 0 new comments

• Comfui images vs Offical demo about Lumina-Image 2.0

  #6741 commented on Mar 4, 2025 • 0 new comments

• No manager present

  #6601 commented on Mar 4, 2025 • 0 new comments

• Get LoraLoader lora name as text

  #561 commented on Mar 4, 2025 • 0 new comments

• ComfyUI/Flux memory utilization when loading model ?

  #4318 commented on Mar 4, 2025 • 0 new comments

• How to edit node id to string not a number for save workflow api

  #4315 commented on Mar 4, 2025 • 0 new comments

• flux1Kijai11GB support

  #4313 commented on Mar 4, 2025 • 0 new comments

• torch.OutOfMemoryError: Allocation on device

  #5278 commented on Mar 4, 2025 • 0 new comments

• Slowed down

  #6950 commented on Mar 4, 2025 • 0 new comments

• Outcome still change just one time after seed set to fixed from random.

  #1382 commented on Mar 4, 2025 • 0 new comments

• ComfyUI crashes due to invalid or unsupported data type Float8_e4m3fn.

  #5335 commented on Mar 4, 2025 • 0 new comments

• HyVideoSampler

  #7012 commented on Mar 5, 2025 • 0 new comments

• VAEDecode Mac BFloat16 is not supported on MPS

  #6254 commented on Mar 5, 2025 • 0 new comments

• ComfyUI not starting, "CUDA error: operation not supported"

  #6843 commented on Mar 5, 2025 • 0 new comments

• ComfyUI get's stuck after clicking queue prompt

  #4300 commented on Mar 5, 2025 • 0 new comments

• How to install comfyUI on macos m3 without a virtual environment? Similar to windows python_embed setup.

  #4299 commented on Mar 5, 2025 • 0 new comments

• After installing the ultralytics library, GPU usage decreased by half and Ksample time doubled

  #4283 commented on Mar 5, 2025 • 0 new comments

• DataLoader.__init__() got an unexpected keyword argument 'in_order'

  #6896 commented on Mar 5, 2025 • 0 new comments

• Latest frontend change breaks stableswarm launch

  #7042 commented on Mar 5, 2025 • 0 new comments

• ComfyUI needs Nvidia GPU even specifying ROCm

  #6649 commented on Mar 5, 2025 • 0 new comments

• Behavior of chained KSampler (adv) change

  #7041 commented on Mar 5, 2025 • 0 new comments

• Force use of VAE decode (tiled)

  #5341 commented on Mar 6, 2025 • 0 new comments

• An error message for the k sampler is displayed

  #5903 commented on Mar 6, 2025 • 0 new comments

• comfyUI can conrtol+A all select nodes, and the author can add the function of reversal, which will be used in some aspects

  #4267 commented on Mar 6, 2025 • 0 new comments

• Initial generation takes an extremely long time (approx. 40 minutes)

  #4266 commented on Mar 6, 2025 • 0 new comments

• Get image preview throught the API

  #2981 commented on Mar 6, 2025 • 0 new comments

• SD3: Error while deserializing header: InvalidHeaderDeserialization

  #3692 commented on Mar 6, 2025 • 0 new comments

• could not create an engine

  #6276 commented on Mar 6, 2025 • 0 new comments

• my cpu is max

  #4262 commented on Mar 7, 2025 • 0 new comments