---
title: "Download Flux Models for ComfyUI Easily with ComfyDL"
date: 2026-01-08T10:00:00.000Z
author: Z.SHINCHVEN
tags: [comfyui, flux, ai-art, tools, python]
canonical: https://atlassc.net/2026/01/08/download-flux-models-for-comfyui-with-comfydl
---
Unlike Stable Diffusion models, which usually come as a single checkpoint file, Flux models consist of a set of model files (UNet, VAE, CLIP, etc.). Downloading them and figuring out where each file belongs in the ComfyUI folder structure can be confusing and tedious.

**ComfyDL** solves this by automating the download and placement process with simple commands.

## Quick Start

Install `comfydl` and set your ComfyUI root path:

```bash
pip install comfydl
comfydl set COMFYUI_ROOT /path/to/your/ComfyUI
```

Download the full Flux.1 suite:

```bash
comfydl flux1
```

Or the FP8 development version:

```bash
comfydl flux1_dev_fp8
```

Done! The tool handles the rest.

## The Challenge with Flux Models

If you are transitioning from Stable Diffusion 1.5 or SDXL to the new state-of-the-art **Flux** models, you might have noticed the installation isn't as straightforward. A functional Flux setup in ComfyUI requires multiple components:

1.  **UNet/Transformer model:** The core diffusion model.
2.  **CLIP/Text Encoders:** Often multiple (e.g., t5xxl_fp16, t5xxl_fp8, clip_l).
3.  **VAE:** The Variational Autoencoder.

Manually downloading these files from Hugging Face and placing them into `models/unet`, `models/clip`, and `models/vae` respectively—while ensuring you have the compatible versions—is error-prone.

## The Solution: ComfyDL

[ComfyDL](https://github.com/ShinChven/comfydl) is a Command Line Interface (CLI) tool designed specifically for ComfyUI users. It manages model sources and destination paths so you don't have to.

### Key Features
*   **Automated Placement:** Downloads files directly to the correct ComfyUI subdirectories.
*   **Preset Model Sets:** Comes with built-in presets for popular models like Flux, Pony, SDXL, and Qwen.
*   **Resume Capability:** Uses `aria2c` (recommended) or `wget` for robust downloads.
*   **Civitai Support:** Can download models directly via version ID or URL.

## Detailed Installation

### Prerequisites
*   Python 3.8+
*   `aria2c` (Highly recommended for faster, resumable downloads) or `wget`.

### Install via pip

```bash
pip install comfydl
```

### Configuration

Tell ComfyDL where your ComfyUI is installed. This ensures files land in the right `models/` folders.

```bash
comfydl set COMFYUI_ROOT /path/to/your/ComfyUI
```

*(Optional) If you download restricted models, you can set your API tokens:*

```bash
comfydl set CIVITAI_TOKEN your_civitai_token
comfydl set HF_TOKEN your_huggingface_token
```

## Downloading Flux Models

Once installed, downloading Flux is as simple as running a single command.

**For the standard Flux.1 set:**

```bash
comfydl flux1
```

**For the optimized Flux.1 Dev FP8 version:**

```bash
comfydl flux1_dev_fp8
```

ComfyDL will fetch all the required text encoders (t5xxl, clip_l), the VAE, and the UNet model, placing them exactly where ComfyUI expects them.

## Other Useful Commands

ComfyDL isn't just for Flux. You can use it for other popular models:

```bash
comfydl pony             # Download Pony Diffusion models
comfydl dreamshaper      # Download DreamShaper
comfydl common_vae       # Download common VAEs
```

You can also browse available models interactively by just running:

```bash
comfydl
```

For more details and updates, check out the [ComfyDL GitHub Repository](https://github.com/ShinChven/comfydl).