📥 Load checkpoint SD1.5 (+Download from CivitAi)

The usual way to try a new SD1.5 checkpoint is a small chore: find it on CivitAI, copy the download URL, wget it into models/checkpoints, restart ComfyUI, then finally select it in a Load Checkpoint node. This node collapses all of that into one step. You browse a curated list of roughly 600 SD1.5 checkpoints right in the node, pick one, and if the file isn't already on disk it pulls it from CivitAI for you - then loads it exactly like a normal checkpoint loader would.

Think of it as a Load Checkpoint node whose dropdown is CivitAI instead of your local folder. That's the whole idea, and it's genuinely handy when you're in a "let me just try that model everyone's posting" mood and don't want to break flow.

How it works

Under the hood it behaves like ComfyUI's core Load Checkpoint. The difference is the list: instead of scanning your models/checkpoints directory, it shows a hand-maintained gallery of CivitAI SD1.5 models. When you run the workflow, the node checks whether that file already exists locally. If not, it downloads it to ComfyUI/models/checkpoints/Bjornulf_civitAI/sd1.5/ using the CivitAI API, then loads it. After the first download the file just sits there, so subsequent runs are instant - and you can point a plain Load Checkpoint node at it too if you'd rather.

The inputs and outputs that matter

• image - this is the model picker. It's named image because it's a thumbnail gallery (each SD1.5 checkpoint shows its CivitAI preview), and there are around 606 entries to scroll through. This is the one input you actually choose.
• civitai_token - your CivitAI API token, grabbed from your account settings on civitai.com. It's only needed for the download; once the file is local you can leave it blank.

Outputs are the three you'd expect from any checkpoint loader - model, clip, and vae - which wire straight into your KSampler, your CLIP Text Encode nodes, and your VAE Decode respectively. You also get name and civitai_url as strings, useful for logging which model produced an image or stamping metadata.

How to install it

Easiest path is ComfyUI Manager: open the Manager, search for Bjornulf_custom_nodes, install, and restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/justUmen/Bjornulf_custom_nodes
pip install -r Bjornulf_custom_nodes/requirements.txt

Then restart ComfyUI. The CivitAI download nodes lean on the civitai-py dependency, which is in that requirements.txt, so don't skip the pip step. On Windows portable installs, run the pip line with the embedded python (.\python.exe -m pip install ...).

Common issues

The download half is where things go sideways. You need a valid civitai_token, and models that are gated, behind early-access, or require you to be logged in on CivitAI can refuse to download even with a token. If a run stalls, the first thing to check is that the token is present and correct.

Early versions of these CivitAI nodes had encoding and link problems on Windows; the author fixed those around v0.71–0.75, so make sure you're on a recent pull. And a reality check on the models themselves: SD1.5 is old, but that ~600-entry list is a fair reflection of how enormous the SD1.5 checkpoint scene got on CivitAI. If you want the newer stuff, there are sibling nodes for SDXL, Pony, and FLUX in the same pack - this one is SD1.5 only.

One more thing worth knowing: this is a huge, very actively developed pack (170-plus nodes, frequent breaking changes in the changelog). Nothing wrong with that, but if a node suddenly behaves differently after an update, check the pack's changelog before assuming you broke something.