ComfyUI-SuperNodes
A miscellaneous node pack of custom implementations needed for workflows released by my SuperComfy YouTube channel.
Nodes (37)
When video and audio shouldn't share one noise schedule
Two model evals per step, two noise knobs, one DiT call — the H3 dual sampler
The boring node that makes DualSamplerCustomAdvanced work
Turn a face detection box into a mask you can actually use
Snap your canvas to a sensible aspect ratio instead of eyeballing it
Crop every video frame around the face — and remember exactly where it was
Crop a face (or anything) to a mask so the model spends its pixels where they matter
Compute exact generation dimensions for any aspect ratio, every time
Paste the fixed face back exactly where it came from
Stitch the processed video crops back without seams or OOM
Track a video object from the middle, not just frame 0
Reserve VRAM from inside the workflow instead of restarting with a flag
How much of your original noise actually survives
Add a denoising step exactly where the schedule needs one
Drop a step you don't need from the noise schedule
Patch a single noise level in your schedule
Recluster your denoising steps without touching the sampler
See your noise schedule before you burn the GPU on it
Put the final denoise step back on a ramp instead of a cliff
Keep your curve, change its range
Brightness, contrast and gamma in one no-frills node
One node to do the whole grade
Kill the color cast a tiled upscale leaves behind
Glue prompt parts together without a stack of text-concat nodes
Split a huge image into same-size tiles for VRAM-starved processing
Rotate the hue wheel without leaving the graph
Paste an image anywhere — even fully off the canvas
Random prompts without learning wildcard syntax
Grayscale that actually follows how your eyes perceive light
Download models straight into the right ComfyUI folder, from inside the graph
Letterbox to an exact canvas — and get a mask of where the padding is
One dropdown to swap a LoRA across every loader
Stitch processed tiles back with feathered seams — even after you upscaled them
The author calls it 'probably useless.' It's actually the part people forget.
Temperature plus tint
A node that intentionally crashes your graph — and why that's useful
Lock the original video in place during SCAIL-2 motion transfer
ComfyUI SuperNodes by SuperCC
[!IMPORTANT] I no longer publish distinct versions to Comfy registry. The latest git commit / nightly version is the only supported version. The 0.3.6 release will quickly be out of date.
Current Nodes
Color
- Adjust Color AIO: Adjust all color options at once
- Adjust Brightness Contrast Gamma: Adjust brightness, contrast, and gamma in one node
- Adjust Hue Saturation: Shift hue and adjust saturation
- Adjust White Balance: Correct white balance via temperature and tint
Extras
- Convert BBox To Mask: Converts bounding box coordinates into a solid mask (requires WanAnimatePreprocess)
Image
- Crop Image using Mask: Crops an image to a mask's bounding box, with optional padding and dimension constraints
- Pad Image Scaled: An alternative method to pad images with adjustable positioning
- Restore Mask Crop: Restores a previously cropped masked region back into its original position
Scheduling
- Sigma Replace: Replaces specific sigma values in a noise schedule
- Sigmas Smoother: Add an intermediary step(s) before full denoise
- Sigmas Graph: Graph the sigma curve
- Sigmas Rescale: Rescales a sigma curve
- Sigma Remove: Remove a sigma at the specified index
- Sigma Insert: Add a sigma at the specified index
Sampling
- DualSamplerCustomAdvanced: Samples a packed audio-video latent with a separate sigma schedule per stream.
- DualSamplerEulerAncestral: Ancestral Euler with separate noise settings for the video and audio streams.
- DualSamplerDPMPP_SDE: DPM-Solver++ SDE with separate noise settings for the video and audio streams.
Tiling
- Create Tiles: Splits an image into an overlapping grid of tiles for tiled processing
- Stitch Tiles: Reassembles processed tiles back into a single image with feathered blending
- Color Match Luminance: Luminance color match
- Luminance Preprocess: Grayscale based on luminance
- Seed VR Calculate Tiles: Probably useless
Tools
- Get Aspect Ratio: Snap to the nearest known aspect ratio
- Image Size Calculator: Calculates target width and height based on aspect ratio
- Model Downloader: Downloads a model from a URL into a selected ComfyUI folder
- Select LoRA Name: Picks a LoRA filename to share across multiple LoRA loaders
- Set Reserve VRAM: Sets --reserve-vram dynamically anywhere in a workflow
- Show Error Message: Halts the ComfyUI pipeline execution with a custom error message
- List Randomizer: Selects a word or phrase from a string list
- Concatenate Multi: Concatenate multiple strings in one node
Video
- Get Video Crop Frames: Crops a frame batch to a uniform size driven by per-frame bboxes or masks (grown to the largest, optionally squared), recording crop metadata for restoration
- Restore Video Crop Frames: Pastes processed crops back onto the background frames using the recorded crop metadata, blending with per-frame masks (crop-space or background-space, auto-detected) or a feathered bbox edge (inner/outer/mid)
- SAM3 Video Track (Masked Index): A modified version of the native video track node that accepts a mask based on a frame index