Ktlyst is a web-based reference canvas for concept artists, illustrators, and visual creators. You gather references on one canvas — drag in images, GIFs, or video frames, or import a whole Pinterest board in one click — and Ktlyst explores new concept directions guided only by the references you chose (a feature called Catalyze). It runs in any modern browser on desktop and tablet, with cloud sync and a free plan.

What is the best PureRef alternative for iPad and tablet?

Ktlyst is a web-based reference board that runs in any browser — including Safari on iPad — with cloud sync so your boards follow you between desktop and tablet automatically. Unlike PureRef, which has no iOS or iPad app, Ktlyst is designed for cross-device creative workflows from the ground up. No installation needed.

Is there a PureRef that works on iPad?

PureRef does not have an official iOS or iPad app. Ktlyst fills that gap: it runs in any browser, syncs your reference boards to the cloud, and works on iPad, tablet, and desktop with the same canvas — so your references are always with you wherever you create.

How does Ktlyst compare to PureRef?

PureRef is a desktop-only overlay tool for keeping reference images visible while you paint or model in other apps. Ktlyst is a cloud-synced, browser-based alternative that adds: iPad and tablet support, Pinterest board import in one click, smart organisation with tagging and collections, freehand sketching overlaid on references, the ability to grab specific frames from videos and GIFs, and a concept-exploration layer (Catalyze) that explores new directions guided by your own references. No installation required — it works in any modern browser.

Do I need to learn prompts or any special tools?

No. You work the way you already do — gather references, sketch, note what you're after. Ktlyst handles the rest behind the scenes, so you stay focused on the vision, not on tooling.

Is this just an image generator?

No. Ktlyst is a reference-driven canvas for clarifying YOUR direction — built around how you already work, collecting references and browsing Pinterest. What it surfaces is an exploration of your own references, for you to take forward.

Can I bring in my Pinterest boards?

Yes. Connect Pinterest and pull an entire board onto a canvas in one move — every pin laid out and ready, instead of saving and re-uploading each one. (Free: 1 board · Pro: unlimited.)

Can I use videos or GIFs as references?

Yes. Drop an MP4, WebM, MOV, M4V, or OGV video — or a GIF — and a frame picker opens so you can scrub through and grab the exact frames you want. Only the frames you pick are saved as reference images; the source video or GIF never leaves your device. (Frame-by-frame GIF picking needs a recent browser; older ones add the first frame instead.)

Can I reuse the same character or element across projects?

Yes. Distil a group of references into one clean, typed element (a character, a costume, a prop), then save it to your Elements library. Drag it onto any canvas to reuse it, merge it with other elements, or paint it into a composition sketch — build it once, use it everywhere.

Can I correct a generated image?

Yes, with Redline (a Pro feature): draw your corrections in red straight on the image — arrows, circles, and handwritten notes — and Ktlyst applies them beside the original, non-destructively, so your original stays put.

Who owns what I create?

Ktlyst claims no ownership over your canvases or the directions you explore — they're yours to take forward into your own work. You're responsible for clearing any rights you need before using output commercially. (How copyright applies to generated images varies by country, so we don't make legal guarantees about that.)

Do you use my images or art to train AI?

No. Ktlyst never uses your uploads, canvases, or prompts to train AI models, and never sells them. Your references and prompts are sent to our AI providers only to generate or analyse images at your request. OpenAI does not use data submitted through its API to train its models; fal.ai processes your content under its own terms to run the generation.

Does it work on tablet?

Yes. Your canvas is cloud-synced, so you can collect on desktop and refine on tablet — or anywhere you work.

What devices and browsers does Ktlyst support?

Editing works on tablets and desktops (iPad, Windows, Mac) in recent versions of Chrome, Edge, Safari and Firefox. Phones get a view-only experience. Ktlyst runs in the browser and needs an internet connection.

Is there a free plan?

Yes — the free plan gives you 5 canvases (each with as many references as you like) and 15 Catalyze runs a month. No credit card required.

July 22, 2026

IP-Adapter vs ControlNet: What's the Difference? (And How to Get Both Without Touching a Single Node)

IP-AdapterControlNetAI image generationno-prompt AI artComfyUI

If you've spent any time reading about Stable Diffusion or ComfyUI, you've almost certainly run into two terms that sound similar but do very different things: IP-Adapter and ControlNet. Both are ways of using a reference image to steer generation instead of typing everything out — but because the names sound alike, it's easy to lose track of what each one actually does.

This article breaks down the difference between IP-Adapter and ControlNet, then looks at how Ktlyst lets you use the same underlying ideas just by arranging reference images on a canvas — no nodes, no settings to memorize.

What IP-Adapter and ControlNet Actually Do

IP-Adapter: turning an image itself into the prompt

IP-Adapter (Image Prompt Adapter) lets you use an image, rather than text, as the prompt. It extracts features from a reference image and feeds them directly into the generation model, which makes it very good at transferring "what to draw" — a character's face, an outfit, the overall style — with much higher fidelity than describing it in words. If you've ever tried to type "silver hair, red eyes, kimono" and gotten something close-but-not-quite, IP-Adapter is built for exactly that gap: showing one image instead of writing a paragraph.

ControlNet: preserving structure, pose, and depth

ControlNet works on a different axis. It extracts structural information — outlines, pose, depth — from a reference image, then keeps that structure intact while a new style or color palette is generated on top of it. The Depth variant in particular is used when you want to redraw a scene's mood or style without breaking its spatial layout. Put simply, ControlNet isn't about "what to draw," it's about "how it's arranged."

So which one is better? Neither — they solve different problems

If IP-Adapter is about bringing in the subject, ControlNet is about building the stage it stands on. In practice, most workflows combine both, depending on the goal — it's less "pick one" and more "understand which job each one is doing."

The catch: using both in ComfyUI takes real setup

Understanding the theory is one thing. Actually wiring up IP-Adapter and ControlNet together in a node-based tool like ComfyUI is another — you're connecting individual nodes, loading the right models, and tuning parameters by hand. "I saw that tangle of wires and gave up" is a common story, and plenty of artists who are genuinely interested in reference-based workflows bounce off before they even get to generate an image, simply because of the setup.

Ktlyst: the same idea, just by placing images in "Main" and "Background"

Ktlyst was designed around the exact two problems IP-Adapter and ControlNet-Depth are trying to solve — transferring a subject faithfully, and preserving a world's structure — without requiring text prompts or node graphs to do it.

The workflow is simple: drop reference images onto a canvas and assign each one a role.

Main images — references for the character or subject you want to draw. Whatever style or characteristics live in these images shape "what" gets drawn.
Background images — references for the setting or space. These control depth and composition — "how" it's all arranged.

Main images answer "how do I render the subject," and Background images answer "how do I build the stage" — the same division of labor that IP-Adapter and ControlNet represent under the hood. You never have to think about model names or nodes; the ordinary act of gathering and arranging reference images, something artists already do, becomes the instruction itself.

Drop in several images, and the shared theme gets picked up automatically

Ktlyst also detects the visual theme running through multiple images placed in the same role. Add a handful of bioluminescent jellyfish photos to a Main role, for instance, and the system reads "glowing creature" as the common thread and carries that theme into future generations. Arranging references is the entire interaction — Ktlyst handles turning that arrangement into language on its own.

How this differs from Midjourney and text-prompt tools

Midjourney, and most AI image tools before it, are built around text prompts as the starting point. Since the quality of what you get depends heavily on prompt-writing skill, artists who don't think in words first often hit a frustrating wall: the image in their head is vivid, but it's hard to translate into a spell-like string of text.

Ktlyst changes that starting point entirely. Instead of "put it into words," the entry point is "gather and arrange." That's not a knock on tools like ComfyUI, which are the right choice when you want maximum technical control — but Ktlyst is aiming at something different: turning the natural act of collecting and laying out reference images into the prompt itself.

The takeaway

IP-Adapter answers "how do I transfer the subject," and ControlNet answers "how do I hold onto the stage." Understanding that split makes it easier to pick the right tool — and to use any of them more effectively.

And if you'd rather experience that division of labor without touching a single setting, tools like Ktlyst — where you simply place images into Main and Background roles — are worth a look. If you'd like to trade the time you spend writing prompts for time spent choosing, arranging, and looking at references instead, give it a try.