🖼️ Isulion Civitai Image Display

The Isulion Civitai Image Display is the second half of the pack's Civitai one-two punch. Its sibling, Isulion Civitai Trending, fetches a text block describing hot images. This node takes that text, downloads the actual images, and hands you real IMAGE tensors you can look at, save, or drop into a pipeline. Text in, pixels out - that's the entire job, and it does it cleanly.

You wouldn't reach for this node on its own; it exists to complete the Trending → Display chain. Without a previous node feeding it, you'd have to hand-paste the image info yourself, which defeats the point. But in the full chain it's the difference between "a node that describes trending images" and "a node that shows me trending images," and the latter is what you actually want for building a style mood board or scouting reference material.

Inputs and outputs

• image_info - a multiline STRING. In the intended flow this comes straight off the Trending node's image_info output. The parser reads it as a set of entries and extracts image URLs, titles, prompts, and model names.
• mode - Single or All. Single grabs the one image at image_index; All pulls every entry in the block.
• image_index - 0-based index into the entries, only meaningful in Single mode.
• target_size - 64 to 2048, default 512, stepped by 64. Each fetched image is resized with Lanczos and padded to a square of this size, centered on black. That's a deliberate choice: you get uniform squares for a tidy preview grid, not your original dimensions.

Outputs are the useful part - it returns more than just pixels:

• image - the IMAGE tensor. Preview it, save it, feed it to a VAE encode for img2img.
• title - the image's title as STRING.
• prompt - the original prompt Civitai has on file for that image. Gold if you're reverse-engineering someone's style, and it's why this node sits in a "Prompt Tools" category.
• image_url - the source URL.
• model - which model produced it.

How it works, and where it's fragile

The mechanism is deliberately simple: it parses the URL out of image_info, does a requests.get, opens the response with Pillow, converts to RGB, and flattens it into a float tensor. On failure - dead link, 404, or an empty input - it doesn't crash; it returns a black error image at target_size so your workflow keeps running and you notice the blank tile instead of a red error. That's friendlier than most nodes, but it means a broken URL looks like "one black square," so glance at the preview before assuming success.

Two caveats. The download happens at graph execution time over the network, so big images or slow links stall the run - keep target_size modest unless you need detail. And because it depends on Civitai's API staying up and the image URLs staying valid, this node is inherently as reliable as the platform feeding it. Civitai's history of churn (see the saga doc) means don't build a critical workflow on its availability.

Install

Standard pack install: ComfyUI Manager → search "Isulion", or clone https://github.com/Isulion/ComfyUI_Isulion into custom_nodes/, then restart. No extra pip packages - it leans on requests and Pillow, both already in ComfyUI.

If you wire Trending → Display and get all black squares, check the Trending node's API key first (missing key = error text, not image entries), then check the image URLs actually resolve in a browser. Those two cover 90% of failures.