ComfyUI
ComfyUI is an open-source, node-graph-based AI image generation workflow tool designed primarily for running and orchestrating diffusion models such as Stable Diffusion. Users can customize image generation pipelines through a visual node-connection interface, supporting text-to-image, image-to-image, model blending, and other operation modes. The tool offers high flexibility and extensibility, making it suitable for researchers and developers with a technical background.
Core Facts
ComfyUI is a node-based visual AI workflow editor that allows users to combine different AI models, preprocessing steps, and post-processing logic together.
ComfyUI is a visual programming framework based on a node graph, where users build generation pipelines by connecting functional nodes, and the community can independently publish custom node packages.
Claude Code manages tool calls through the MCP (Model Context Protocol) protocol.
Timeline (last 90 days)
safetensors can be opened in transformers, native vLLM environments, ComfyUI, and Apple's MLX toolchain.
Format choice is determined by the runtime already in use: Ollama/LM Studio/llama-server automatically select GGUF, Python/transformers pipelines automatically select safetensors, vLLM on GPU servers natively supports safetensors, and ComfyUI or image pipelines use safetensors.
By default, after generating one image, the SDXL checkpoint, CLIP text encoder, and VAE remain resident in VRAM, occupying about 8GB; after calling /free, VRAM falls back to a baseline of about 2.2GB.
ComfyUI provides a /free endpoint that can actively unload loaded model weights, clear cached tensors in VRAM, and trigger Python garbage collection.
In ComfyUI, running SDXL alone to generate one image takes about 15 to 30 seconds.
In the Mac version of Locally Uncensored, image generation does not depend on ComfyUI but instead runs directly on the machine using an engine installed by the app itself; the Windows/Linux versions drive image generation through ComfyUI.
In Locally Uncensored 3.0.6, Qwen-Image 2.1 LoRAs that come with an alpha now compute weights the ComfyUI way (alpha divided by rank), fixing the previous problem of running at the full configured strength that made the effect too strong.
Running a ComfyUI custom node graph locally with models such as Wan 2.1 or Wan 3.0 requires more than 24GB of VRAM (NVIDIA RTX 3090/4090 or a cloud A100 instance).
In the ComfyUI ecosystem, Wan 3.0 usually serves as the core sampling node, working together with auxiliary nodes such as motion modules, depth guidance, and face locking to form a complete video synthesis pipeline.
Unsloth now directly reuses the text encoder already present in the ComfyUI folder instead of downloading another copy.
40 more timeline events
All Facts (20)
ComfyUI is a node-based visual AI workflow editor that allows users to combine different AI models, preprocessing steps, and post-processing logic together.
80%VerifiedComfyUI is a visual programming framework based on a node graph, where users build generation pipelines by connecting functional nodes, and the community can independently publish custom node packages.
80%VerifiedClaude Code manages tool calls through the MCP (Model Context Protocol) protocol.
80%VerifiedComfyUI是一个基于节点式工作流的Stable Diffusion图形界面工具，用户可以通过拖拽和连接不同功能节点来构建自定义的图像处理流程
80%VerifiedThe FastVideo project focuses on inference acceleration for video generation, with the goal of making video generation faster and more resource-efficient.
75%Verified生成一段8秒、24帧/秒的视频意味着模型需要在时间轴上同时保持192帧画面的语义一致性，计算量约为同分辨率单张图像的数十倍
75%VerifiedComfyUI is an open-source node-based workflow tool released by developer comfyanonymous in early 2023.
75%VerifiedComfyUI is a node-based workflow interface built on Stable Diffusion.
70%VerifiedComfyUI uses a Directed Acyclic Graph (DAG) architecture, where users build processing pipelines by connecting nodes with different functions.
70%VerifiedComfyUI Assistant is a custom extension that embeds an AI assistant into the ComfyUI interface and can read and manipulate the user's current node graph.
65%VerifiedThe video frame interpolation workflow can increase the frame rate from 24 fps to 48 fps.
65%VerifiedThis workflow can generate continuous videos up to 60 seconds long on a 12 GB GPU.
65%VerifiedComfyUI is a node-based Stable Diffusion graphical user interface that adopts a dataflow programming paradigm, decomposing the generation process into a directed acyclic graph (DAG).
65%VerifiedUnlike the form-filling interaction style of Automatic1111 WebUI, ComfyUI's node-based design is more flexible and supports complex conditional branching and multi-stage generation workflows.
65%VerifiedThe rise of workflow orchestration tools such as ComfyUI and Dify provides the infrastructure support for the trend of multi-model collaborative AI creation.
65%VerifiedAfter integration into ComfyUI, MiniMax H3 supports six core workflows: text-to-video, image-to-video, first-and-last-frame animation, reference-image-to-video, audio synchronization, and image editing.
65%VerifiedComfyUI是开源图形化工作流工具，其核心设计理念是将生成管线拆解为可视化的节点图
65%VerifiedComfyUI is centered on a node-based workflow that offers strong flexibility but has a steep learning curve for beginners.
65%VerifiedThe higher the reasoning effort level, the stronger the reasoning capability but the faster token consumption, representing a trade-off between cost and intelligence.
65%VerifiedNode-based workflow editors such as ComfyUI allow creators to chain models by dragging and dropping nodes without writing code, and to package and share workflows in JSON format.
65%