Model Downloader
Download models into the right folders without leaving ComfyUI
- status_message
The single most annoying part of ComfyUI is getting model files into the right directory - drop a GGUF in the wrong subfolder and your dropdown is empty, and there's no error message to help. ModelDownloader exists to kill that whole class of problem: you give it a folder name and URLs, and it downloads straight into the correct spot under ComfyUI/models/. It's the node version of "download this, save it here," except the "here" is guaranteed to be right.
How it works
You type a list with a simple format - a folder name line, then the URLs that go in it:
diffusion_models
https://huggingface.co/unsloth/Qwen-Image-2512-GGUF/resolve/main/qwen-image-2512-Q8_0.gguf
clip
https://huggingface.co/unsloth/Qwen2.5-VL-7B-Instruct-GGUF/resolve/main/Qwen2.5-VL-7B-Instruct-Q8_0.gguf
Each block maps to a subfolder under models/, and the node downloads in parallel with a configurable number of worker threads. It also checks what's already there, skipping complete files and picking up interrupted ones. Two download paths, toggled by checkboxes:
use_hf_download- off by default, which uses plain HTTP chunked downloads (maximum compatibility, per the author's tooltip). On, it useshuggingface_hub- files stage in the HF cache first, then move into your models folder automatically.use_s3c- pairs with HF mode and uses thes3impleclientaccelerator for faster pulls. Only works whenuse_hf_downloadis on, and only if you've installeds3impleclient.
You can also skip typing entirely: model_template has ready-made download lists for the pack's recommended models - Qwen-Edit bundles, Z-Image Turbo in GGUF/FP8/FP16 variants with VRAM tiers in the names, and an Ultraflux VAE. Selecting one overrides your manual list, which the tooltip warns you about.
Inputs and output
download_list- the folder/URL blocks above.chunk_size_mb(default 10) andmax_workers(default 8) - tune for your connection.model_template- the one-click presets.- The rest are the two download toggles.
Output is a single status_message string telling you what got downloaded, skipped, or failed - useful for glancing at in a Show Text node.
Installing it
cd ComfyUI/custom_nodes
git clone https://github.com/dseditor/ComfyUI-ListHelper
Restart ComfyUI; find it under ListHelper/Tools. The huggingface_hub and s3impleclient packages aren't in the pack's requirements, so they install on demand only if you enable those modes - plain HTTP mode needs nothing extra. ComfyUI Manager: search "ComfyUI-ListHelper".
Where people get burned
The URL format matters. These need direct resolve/main/… file URLs from HuggingFace, not repo pages - copy the "download" link, not the model page URL. And a word of caution from the wider ecosystem's habit of downloading workflow JSONs: only run this with URLs you actually trust. A malicious node pack is how a lot of people got burned in this ecosystem, and a "model downloader" is a great place for bad actors to hide. For the legit Qwen-Edit / Z-Image templates it ships with, it's genuinely convenient - just keep your own list to repos you recognize.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| download_list | STRING | diffusion_models https://huggingface.co/unsloth/Qwen-Image-2512-GGUF/resolve/main/qwen-image-2512-Q8_0.gguf CLIP https://huggingface.co/unsloth/Qwen2.5-VL-7B-Instruct-GGUF/resolve/main/Qwen2.5-VL-7B-Instruct-Q8_0.gguf | 下載列表格式: 資料夾名稱 URL1 URL2 ... 注意:若選擇了模型範本,此輸入將被忽略。 |
| use_s3c | BOOLEAN | false | 使用 S3impleClient 加速下載(需安裝 s3impleclient)。 僅在開啟 HF 下載時有效。 |
| use_hf_download | BOOLEAN | false | 使用 HuggingFace Hub 下載。 關閉時使用普通 HTTP 下載(相容性最高)。 開啟時檔案會先下載到 HF 快取,再自動搬移到 ComfyUI models 資料夾。 |
| chunk_size_mb | INT | 101–128 | 每次 HTTP 請求的區塊大小 (MB) |
| max_workers | INT | 81–32 | 最大平行下載執行緒數 |
| model_template | COMBO | 無 | 選擇模型範本。選擇範本後將使用範本中的下載列表,忽略上方手動輸入的內容。 選擇「無」時使用上方手動輸入的下載列表。 |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| status_message | STRING | — |