Gigapixel Upscale Settings

Another settings-only node in the pack: it bundles the enhancement options Topaz normally hides in its desktop GUI into one object you wire into the GigapixelAI node's upscale input. The main node runs fine without it (defaults apply), but if you've ever wished you could dial in denoise or artifact removal inside your workflow instead of eyeballing it in a separate app, this is the wire.

The inputs, all of which map to knobs you'd find in the Topaz app:

• enabled - master switch for these settings. Off means the sliders are ignored.
• sharpen (0–100, default 1) - sharpening strength. Note the tiny default: your source is usually a freshly rendered AI image, not a soft photograph, so Topaz's sharpen should mostly stay out of the way.
• denoise (0–100, default 1) - same story. Diffusion output isn't noisy in the way a camera ISO 6400 frame is.
• compression (0–100, default 67) - compression-artifact removal, and the one you'll actually feel. Feed it a JPEG-damaged or heavily compressed source and it smooths the blockiness.
• fr (0–100, default 50) - face recovery strength. The option to reach for on portraits, and the main reason people call Topaz out as better on faces than the free ESRGAN models.
• pre_downscaling (50–100, default 75) - Topaz pre-downscales the source before processing to speed things up. Lower percentage is faster but surrenders detail; higher keeps more of the source.

Output: upscale_settings, which connects to the GigapixelAI node's upscale input. That's all it does.

The honest take, matching the upscaling community's consensus: for freshly generated AI images you mostly care about compression and fr - sharpen and denoise near zero is usually right, because the render is already clean. This node becomes the whole point of the pack when you're doing photo restoration or upscaling heavily compressed sources, where artifact removal and face recovery do the real work. That's exactly the job the open-source pixel rungs (plain ESRGAN) are bad at and why a paid baseline like Topaz keeps its fans.

Install and gotchas are the pack's: you need a licensed Topaz Gigapixel 8 install, the pack is Windows-only, and GigapixelAI needs the right gigapixel_exe path. And same as its model-settings sibling: wiring this up does nothing until it's connected to the main node's upscale input.