Dickson_TTP_Preprocessor_cufoff

Let me save you a confusing hour. Of the three tile preprocessors in this pack, this is the one you should skip. Dickson_TTP_Preprocessor_cufoff - yes, "cufoff", the typo is in the name - is meant to be a low-pass-frequency variant for TTPlanet's SDXL tile ControlNet, but the shipped code is broken: it calls a function called apply_low_pass_filter that doesn't exist anywhere in the pack. Run it and you get a NameError, your workflow dies mid-graph, and there's no setting that fixes it.

That's not me guessing from a stale README. I checked the source. core.py calls the function, and the pack's dicksonttplanetcontrolnet.py module defines only apply_gaussian_blur and apply_guided_filter. The name isn't even imported. It looks like a half-finished experiment that got published anyway - which tracks, because the same changelog that renamed these nodes to avoid colliding with TTPlanet's own pack (v1.0.7) never got around to finishing this one.

What it was supposed to do

In the same family as its siblings, the idea was: preprocess your input image into the smoothed, flattened conditioning image that a tile ControlNet wants, so the model regenerates fresh detail instead of copying pixel noise. The intended pipeline was a low-pass filter controlled by cutoff_frequency (0–256, default 100) and filter_strength (0.1–10, default 1), then a downscale-and-upscale pass via scale_factor, then a Gaussian blur via blur_strength. Inputs: image, scale_factor, blur_strength, cutoff_frequency, filter_strength. Output: image_output, meant to wire into a ControlNet Apply node alongside the TTPlanet SDXL tile model.

What to do instead

Use Dickson_TTP_Preprocessor_Simple or Dickson_TTP_Tile_Preprocessor_GF from the same pack - both work, both do the smoothing job, and both are listed here with their own pages. If you specifically need frequency-cutoff control, this node is the wrong place to look; grab a proper low-pass or Fourier filter node from a maintained pack.

The install is otherwise identical to the rest of the pack - ComfyUI Manager, search dicksondickson, or git clone https://github.com/dicksondickson/ComfyUI-Dickson-Nodes into custom_nodes, then restart - and you still need the TTPlanet tile ControlNet model in ComfyUI/models/controlnet for any of these to be useful. But for this node specifically, the troubleshooting answer is shorter: don't. It's a trap dressed as a variant.