FF: Pixel Boost

Let's be upfront about this one: it's the least-documented node in the pack. The README's entire description of it is one line under "Detection & Tools" - "Configure pixel boost settings" - and it never shows up in any of the example workflow diagrams. Every other swap node in this pack (SwapFaceImage, AdvancedSwapFaceImage, FaceSwapApplier, and the video equivalents) already has pixel_boost baked in as its own dropdown, so the honest question is: what's a standalone node for, if the setting already lives everywhere you'd use it?

Here's what the schema actually tells us, and the best read of it. Pixel boost, per the pack's own docs, is the resolution the swapper model processes the face crop at internally - 256x256 is fast and basic, 1024x1024 is slow and best-quality, 512x512 is the recommended middle. This node takes an image and the same four-option pixel_boost choice everything else has, and returns the image back unchanged alongside a pixel_boost_setting output - a plain STRING, not a resolution or a resized image. No other node in the pack accepts a STRING for pixel_boost, so this doesn't look like it's meant to feed the setting back into a swap node directly.

The most plausible use: pick your pixel boost tier once, and get it back as text you can route somewhere that wants a label rather than a widget value - a Show Text node for on-canvas documentation, or baked into a filename so your saved outputs record which resolution tier they were rendered at. The image passthrough is likely there so the node can sit inline in a chain without breaking anything wired around it, rather than because it performs any resizing of its own - the actual pixel-boost resize happens inside the swapper models, not here.

Inputs and outputs that matter

• image (IMAGE) - passed straight through.
• pixel_boost (default 512x512, 4 choices: 256x256, 512x512, 768x768, 1024x1024) - the only real setting on the node.

Outputs: image (IMAGE, unchanged) and pixel_boost_setting (STRING, the label of whatever you picked).

If you're not sure whether you need this node: you probably don't. Just set pixel_boost directly on whichever swap node you're actually using - AdvancedSwapFaceImage, AdvancedSwapFaceVideo, or FaceSwapApplier all have the field built in.

Installing it

ComfyUI Manager: search Facefusion_comfyui, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/huygiatrng/Facefusion_comfyui.git
cd Facefusion_comfyui
pip install -r requirements.txt

Full restart afterward. This particular node needs no model download of its own - it's pure configuration, no ONNX involved.

Common issues

There isn't much of a track record to draw on here - this is the pack's least-used node by a wide margin, and neither the README nor its own launch thread discuss it in any detail. The one thing worth flagging: if you expected the output image to actually look different (sharper, resized, upscaled), it won't - this node doesn't perform the boost itself, it just carries the setting. For anything install-related - nodes missing after setup, an ONNX runtime error on a different node in the same graph - the fix is the same as everywhere else in the pack: make sure pip install -r requirements.txt ran inside ComfyUI's own Python environment, then fully restart rather than refreshing the browser.