Exactly Illustrative (custom model)

The name is doing double duty. RunwareArch_exactly_illustrative is the Runware pack's way of saying "any Exactly Illustrative–architecture checkpoint in the cloud catalog, as a node." You pick the specific model, you get an IMAGE tensor back, and you never download a weight file. That's the whole deal with this pack: every node is a thin remote call wearing a ComfyUI costume.

Exactly Illustrative is an art-style checkpoint tuned for clean, painterly illustration output. If you've been eyeing it but your card can't hold an SDXL-era model comfortably - or you just don't want to manage another models/checkpoints download - this node is the zero-install path. You pay per run instead of per gigabyte.

How it works

The node builds a request dict (taskType: imageInference, plus your model's AIR and widget values) and sends it over the Runware REST API via their Python SDK. The image comes back as a URL, which the node downloads and converts to a native IMAGE tensor. Wire that into Preview Image or Save Image and you're done. There's no local diffusion loop anywhere in this pack - the only compute you spend is the request.

Inputs that matter

• model - the AIR (model identifier) of the exact checkpoint you want. It defaults to empty here, so use the search button next to the field to browse the Runware catalog for this architecture and pick one. This is the one input you genuinely can't skip.
• positivePrompt - your prompt, plain English.
• width / height - note the floor: both are 1024–2048 in 64px steps. This is an art model that doesn't want to go small.
• settings.quality - low or high. Low is the budget default; high costs more but is worth it if the output is going anywhere public.
• referenceImages - an optional typed socket; feed it a style/character reference from a Runware/Params builder if the checkpoint supports it.
• numberResults - crank past 1 and each result uses a different seed, so it's an instant variation batch.

Install

Same as every node in this pack:

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

Or skip all that and use ComfyUI Manager → search "Runware" → Install → restart. Then grab an API key from the Runware dashboard and paste it into ComfyUI Settings → Runware API key (or set RUNWARE_API_KEY). No key, no runs - the node errors out with "No Runware API key."

Gotchas

• Runware is a prepaid top-up API. There's a minimum top-up and it wants a card linked, which a lot of people find off-putting, but per-image cost ends up about as cheap as cloud inference gets.
• If nothing appears after queueing, you almost certainly forgot to wire the image output into a Preview/Save node.
• outputFormat defaults to JPG; switch to PNG if you want lossless, and WEBP if size matters.

The one-line pitch: a GPU-free way to run an art checkpoint whose weights you'd rather not manage. Cheap, fast, and it slots straight into your existing ComfyUI graph.