Superside Topaz Upscale Image

Topaz is the desktop-app name everyone knows for cleanup upscaling - the "make this low-res image look good" product. This node brings the same model family to ComfyUI via fal.ai's topaz/upscale/image endpoint, with all ten model variants exposed as a dropdown. If you've ever wanted Topaz's particular flavor of enhancement inside a graph instead of a separate app round-trip, this is it.

What it does

It runs a Topaz enhancement model on your image. Where most upscalers give you one behavior, Topaz gives you a menu of them, and the model dropdown is where the personality lives: Low Resolution V2 for genuinely tiny sources, Standard V2 (the default), CGI for renders, High Fidelity V2 for keeping things accurate, Recovery/Recovery V2 for damaged photos, Redefine for creative re-rendering, Wonder, Text Refine, and Standard MAX. Picking the right one is most of the craft here - the wrong variant on the wrong source is how you get an over-smoothed mess.

The inputs that matter

You won't touch most of these, so here's the short list:

• model - the upscale style, default Standard V2. The one that matters most.
• upscale_factor - 1–4, default 2.
• face_enhancement (default ON) with face_enhancement_creativity (0) and face_enhancement_strength (0.8) - Topaz's face-aware cleanup. If faces come back plastic, drop the strength.
• sharpen, denoise, fix_compression - all 0–1, default 0.5. The cleanup trinity.
• subject_detection - All / Foreground / Background, controls what gets processed.
• strength (default 0.5), creativity (1–6), texture (1–5), prompt + autoprompt, detail (0.5) - the deeper enhancement knobs; prompt lets you nudge the creative models with text, and autoprompt (default ON) generates one for you.
• crop_to_fill (default OFF) and output_format (jpeg default / png).

Output: a single IMAGE - the upscaled result.

Installing it

Part of comfyui-superside-nodes - clone, install requirements, restart:

cd ComfyUI/custom_nodes
git clone https://github.com/Superside/comfyui-superside-nodes.git
cd comfyui-superside-nodes
pip install -r requirements.txt

Paste your fal api_key. No model files; the enhancement runs on fal's servers and is metered per call.

The honest take

Where this fits in the upscaling ladder: Topaz is cleanup upscaling - it sharpens, denoises, and reconstructs, but it's not the detail-inventing restoration of a SeedVR2, and it won't add fabric weave the way a diffusion upscaler can. Use it for "clean this image up and make it bigger" with a specific model pick; use SeedVR2 when you want plausible new detail. And because this is the hosted version, the good news is there's no Topaz desktop license to juggle - the bad news is every run is metered, so the per-call cost of those ten variants is something to keep an eye on during batch runs. Dial the model, then dial sharpen/denoise conservatively; Topaz's defaults lean toward "safe," which is usually right for a first pass.