Introduction

Two new partner nodes just landed in ComfyUI, and they cover two jobs that usually send you out to a browser tab. Kling Virtual Try-On dresses a person photo in a garment photo, and Clarity AI Crystal Upscale pushes an image up to 100 megapixels while keeping faces and skin honest. Both are closed, hosted models that run as API nodes through Comfy credits, and both merged into ComfyUI master on October 7 (UTC), so you'll find them on nightly today and in the next tagged release.

Here's what each one does, what it costs per run, and where it fits in a real graph.

Kling Virtual Try-On node

The node is called KlingTryOnNode (display name Kling Virtual Try-On) and sits under partner/image/Kling. It came in through ComfyUI PR #16844 and talks to Kling's current solutions/virtual_try_on endpoint, not the older Kolors try-on route.

Inputs

  • person_image: one person, ideally front-facing or three-quarter view.
  • garment_image: a product shot, flat lay, mannequin, or on-model photo. An on-model photo can carry the rest of that outfit across too.
  • keep_pose (advanced, on by default): turn it off and Kling may re-pose the person for a better outfit presentation.
  • seed: only controls re-runs; results stay non-deterministic either way.

Limits and price

  • Clothing only. Shoes, bags, and accessories aren't supported.
  • Both images get downscaled to 2048px on the longest side before upload.
  • The output matches the person image's size, capped at 2048px.
  • The node's price badge reads $0.16 per run.

Clarity AI Crystal Upscale node

ClarityCrystalUpscaleNode (display name Clarity AI Crystal Upscale) arrived in PR #16817 under partner/image/Clarity AI. It calls Clarity AI's Crystal Upscaler, which Comfy describes as a high-fidelity upscaler that stays faithful to the source and restores faces, skin, and fine texture.

Inputs

  • image: exactly one image. Batches throw an error.
  • scale_factor: 1.0 to 200, default 2.0.
  • creativity: a 0 to 10 slider, default 0. Higher values let the model rebuild more detail instead of strictly preserving the original. It does nothing on images whose short side is 256px or less.

Limits and price

  • Output caps at 100 megapixels and 65,535px per side. The node checks this before it sends anything, so an oversized request fails fast instead of burning credits.
  • Results come back as PNG.
  • The price badge reads about $0.023 per megapixel.

Why these two belong together

Put them side by side and you've got a tidy e-commerce or lookbook chain inside one graph: generate or load a model shot, swap the outfit with Kling Try-On, then run Crystal Upscale for a print-ready or zoomable product image. Everything stays in ComfyUI, so you can batch garments through the same person image without juggling three web apps.

A few practical notes:

  • Leave creativity at 0 for product work. Fabric patterns and logos on the garment should survive the upscale untouched.
  • If the try-on result looks stiff, flip keep_pose off and compare.
  • Both are API nodes, so you'll need to be signed in to a Comfy account with credits.

How to get them

Update ComfyUI to the latest master or nightly build, then search the node library for "Kling Virtual Try-On" or "Crystal Upscale".

cd ComfyUI
git pull
pip install -r requirements.txt

Desktop and portable users can switch to the nightly channel or wait for the next tagged release.

Original Source

Conclusion

These aren't flashy new foundation models, but they're the kind of quiet plumbing that saves you hours. Try-on plus a faithful upscaler in one graph is a real workflow for anyone making fashion, merch, or lookbook content, and both nodes are live for nightly users right now.

—Aurelia ♡