Navify AINavify AI
概要ツールを送信
ComfyUI

ComfyUI

Build visible, adjustable generative AI workflows by connecting nodes on a canvas

画像生成Node WorkflowOpen SourceLocal InferenceTrial

ComfyUI is an open-source, node-based interface and inference engine for generative AI. It runs locally on Windows, macOS, and Linux, lets users connect nodes into visible workflows, and offers a cloud option with paid plans and an API for production use.

公式サイトへ

更新日

2026-10-21

対応プラットフォーム

Windows / macOS / Linux / Web

お試し

Freemium: ComfyUI is open source and free to run locally. Comfy Cloud offers 5 free runs on real GPUs plus free Comfy Agent credits for new users; paid plans include Standard at $16 per month billed yearly ($20 monthly), Creator at $28 ($35 monthly), Pro at $80 ($100 monthly), Team at $630 ($700 monthly), and custom Enterprise. Cloud usage is charged by active GPU time and credits; confirm current plans on the official pricing page.

* 料金は参考情報です。最新の料金は公式サイトをご確認ください

ComfyUI
comfy.org

ComfyUIの主な機能

  • Node Workflows

    Connect nodes on a canvas and adjust every model, parameter, and processing step in the graph.

  • Local Inference

    Install ComfyUI on your own machine and keep models, workflows, and outputs on local hardware.

  • Cloud and API

    Use hosted GPUs and credits in Comfy Cloud, or deploy a workflow as an API through the developer platform.

  • Comfy Agent

    Describe the result you want and let Comfy Agent plan, build, and run the workflow on the canvas.

ComfyUIのおすすめ用途

Visual AI workflow designers

Inspect every model, parameter, and processing step in a node graph instead of relying on a fixed one-click pipeline.

Local model users with their own GPUs

Run ComfyUI on Windows, macOS, or Linux and keep models and data on their own hardware.

Teams moving workflows into production

Deploy a ComfyUI workflow as an API or use Comfy Cloud to scale it on hosted GPUs.

長所

  • Open source and free to run locally
  • Every model, parameter, and processing step is visible and adjustable
  • Large node and workflow ecosystem with community templates
  • Cloud option and API for production deployment

注意点

  • Local use assumes your own hardware and model setup
  • Cloud runs are billed by active GPU time and credits
  • Node graphs have a steeper learning curve than one-click generators

ComfyUI のクイックスタートと実践のコツ

クイックスタート

  1. 1

    Start with a workflow that already matches your task instead of an empty canvas, then delete nodes you do not need: every remaining node stays an explicit parameter you can inspect.

  2. 2

    Run one image with a fixed seed before touching any other setting, so later changes can be attributed to a single variable.

  3. 3

    When the workflow behaves as intended, use Comfy Agent to describe the goal and compare its graph with yours; it is faster for scaffolding than for fine control.

実践のコツ

The usual mistake is rebuilding a graph every session and losing the good one. Save finished workflows as JSON and keep a fixed seed plus one quality preset as your baseline, then branch from it with duplicate-and-tweak instead of editing in place. On a local GPU, watch VRAM before node count: out-of-memory errors usually come from resolution and batch size, not from a missing node.

ComfyUIの代替ツール

このツールが合わない場合は、次の代替候補をご覧ください