comfyui-ysnodes Introduction
comfyui-ysnodes is a collection of nine custom nodes for ComfyUI. It brings common workflow tasks into a compact toolkit: choosing an image size, preparing images and masks, creating sampling schedules, using reference images, and receiving a notification when a workflow finishes.
The pack is designed to work without other custom node packs. Its nodes use ComfyUI’s built-in capabilities rather than requiring another collection of nodes. You can find them under the YSNodes/ categories in ComfyUI’s node menu.
For an AI artist, the practical benefit is fewer repeated steps. For example, you can choose a portrait aspect ratio and a megapixel target instead of looking through a long list of fixed resolutions. You can also resize a reference image and prepare it for a compatible workflow without assembling as many separate utility nodes.
How comfyui-ysnodes Works
A ComfyUI workflow passes information from one node to the next through connections. comfyui-ysnodes provides nodes that prepare that information at different stages:
- Choose a canvas size. A Size Picker calculates a width and height from an aspect ratio and megapixel target, or uses the dimensions of a connected image.
- Prepare an image when needed. Image-scaling nodes resize images and optional masks. Some can also use a VAE to turn an image into a latent, the representation used by image-generation models during sampling.
- Guide generation. Sigma Generator provides a sampler and sampling schedule. Reference Conditioning adds an encoded reference image to conditioning for workflows that support that approach.
- Finish the workflow. Purge & Notification can clear caches and request a browser sound or operating-system notification.
Think of pixel images as the pictures you view and edit, and latents as the working format a generation model uses internally. Nodes that output only an image are useful when the next step needs visible pixels. Nodes that output a latent are useful when the next step samples from, or otherwise processes, that latent.
Not every node belongs in every workflow. You can use a Size Picker alone for text-to-image generation, or combine an image loader, scaling node, and VAE-based output for an image-to-image workflow.
comfyui-ysnodes Features
Sigma Generator
Category: YSNodes/sampling
Sigma Generator combines the roles of KSamplerSelect, BasicScheduler, and SplitSigmas. It takes a model, sampler_name, scheduler, steps, skip_start_step, and denoise, then outputs SIGMAS and a SAMPLER for a SamplerCustom workflow.
The sampler and scheduler determine how sampling is carried out. steps sets the schedule length, while skip_start_step and denoise let you adjust which portion of that schedule is used. This node is useful when you want those related controls together rather than spread across several nodes.
Size Picker
Category: YSNodes/utility
Size Picker calculates image dimensions and creates an empty latent for generation. It offers two ways to choose the size:
- Computed size: Select an
aspect_ratioandmegapixelstarget. Themultiplesetting rounds the dimensions to a chosen increment. - Image-driven size: Connect an optional
imageand enableuse_image_size. Withfit_megapixels, you can choose whether the connected image keeps its exact size or is resized toward the megapixel target while preserving its aspect ratio.
Other inputs include width_override, height_override, batch_size, and latent_type. The outputs are image, width, height, batch, and latent. When no image is connected in computed-size mode, the image output is a blank placeholder at the calculated size.
The 17 available aspect ratios are 1:1, 4:5, 5:7, 2:3, 3:4, 5:12, 9:16, 9:21, 1:2, 5:4, 7:5, 3:2, 4:3, 12:5, 16:9, 21:9, and 2:1. For example, 9:16 suits a tall composition, while 16:9 suits a wide one.
Size Picker Full
Category: YSNodes/utility
Size Picker Full offers the same sizing controls as Size Picker, plus optional VAE encoding. It outputs latent_empty for starting a new image and latent_from_image for a workflow that begins with a connected image, such as image-to-image generation or inpainting. It also outputs image, width, height, and batch.
Choose Size Picker when you only need an empty starting latent. Choose Size Picker Full when you also need an image encoded into a latent. Connect a suitable vae when using the image-encoding output.
Reference Conditioning
Category: YSNodes/conditioning
Reference Conditioning encodes an image and attaches its latent to conditioning for compatible reference-image workflows, including Flux Kontext-style workflows. Its inputs include conditioning, pixels, vae, upscale_method, megapixels, multiple, enable_megapixels, and an optional mask. It outputs conditioning, image, mask, and latent.
Enable megapixel scaling when you want to adjust the reference image’s size before encoding it. Keep in mind that attaching a reference latent is a workflow-specific operation: the rest of your workflow must support that form of conditioning.
Redraw Size Calculator
Category: YSNodes/image
Redraw Size Calculator resizes an image within a minimum and maximum megapixel range. You can select from 14 size presets spanning SD 1.5 through 4K UHD, use min_override and max_override, choose an upscale_method, and connect an optional mask.
It returns both resized and original image and mask outputs, their respective dimensions, and target_megapixels. Keeping both versions available makes it easier to use one for processing while retaining the source image elsewhere in your workflow.
Purge & Notification
Category: YSNodes/utility
Purge & Notification is an end-of-workflow utility. It can purge the VRAM cache, optionally purge models, and request a browser sound and operating-system notification. Its controls include purge_cache, purge_models, alert_mode, system_notification, notification_text, play_sound, sound_volume, and sound_file. An optional signal input and signal output let you place it in a workflow chain.
Place it after the last Save Image node when you want an alert that the workflow has reached its end. For a local notification sound, place a browser-playable file such as an MP3 or WAV in comfyui-ysnodes/web/assets/ and set sound_file accordingly. The source README also notes support for a browser-playable URL.
ImageScaleToTotalPixels
Category: YSNodes/image
ImageScaleToTotalPixels scales an image and optional mask toward a target megapixels value and can encode the result with a vae. Its inputs include pixels, upscale_method, megapixels, enable_megapixels, optional mask, and optional vae. It outputs image, latent_image, mask, and latent_mask.
Use it when a later step needs both resized pixels and latent outputs. The enable_megapixels switch lets you control whether megapixel-based resizing is applied.
ImageScaleToNormPixels
Category: YSNodes/image
ImageScaleToNormPixels resizes an image and optional mask without VAE encoding. It provides scale_by as well as megapixels, plus upscale_method, multiple, and enable_megapixels. It outputs image and mask.
Use it when the next node needs a resized image or mask, but not a latent. With scale_by=1.0 and enable_megapixels=false, it acts as a passthrough. Unlike ImageScaleToTotalPixels, it does not produce latent outputs.
Load Image Temporarily
Category: YSNodes/image
Load Image Temporarily loads an image from ComfyUI’s /temp folder. Its image control includes a dropdown and an upload button. It outputs image, mask, width, and height.
This is useful for a short-lived reference or test image that you do not want to add to ComfyUI’s regular /input folder.
comfyui-ysnodes Model Compatibility
comfyui-ysnodes does not provide image-generation models. Instead, Size Picker and Size Picker Full provide a latent_type setting that shapes their latent output for the model architecture used elsewhere in your workflow.
| Model used in your workflow | Select latent_type | Latent format listed by the author |
|---|---|---|
| SD 1.5, SD 2.x, or SDXL | SD / SDXL | 4 channels, dimensions divided by 8 |
| SD3 or AuraFlow | SD3 / AuraFlow | 16 channels, dimensions divided by 8 |
| Standard Flux | Flux | 16 channels, dimensions divided by 8 |
| Flux 2 | Flux2 | 128 channels, dimensions divided by 16 |
Choose the option that matches your generation model. For example, selecting Flux2 for a Flux 2 workflow gives the Size Picker a different latent shape than selecting SD / SDXL. A mismatch can lead to dimension errors rather than a different artistic style.
What's New with comfyui-ysnodes
Version 1.2.0
- Added ImageScaleToTotalPixels for megapixel-based image and mask scaling with VAE-encoded outputs.
- Added ImageScaleToNormPixels for scaling images and masks without VAE encoding.
- Added Load Image Temporarily and its
/tempupload button. - Added
latent_typeto both Size Pickers so their latent output can match SD/SDXL, SD3/AuraFlow, Flux, or Flux 2 workflows. This also fixes the earlier behavior in which the latent output always used the SD/SDXL format. - Added
fit_megapixelsto control whether a connected image retains its exact size or is fitted to the megapixel target. - Added an
imageoutput to both Size Pickers, providing either the connected image or a blank placeholder at the computed size.
Version 1.1.0
- Replaced the Size Picker’s 35 fixed resolution presets with a system based on 17 aspect ratios, a megapixel target, and a rounding multiple. This changes how older Size Picker workflows are configured.
- Updated Size Picker Full to use the same sizing system and provide separate
latent_emptyandlatent_from_imageoutputs. Older workflows using its previous configuration may need attention. - Added Reference Conditioning for compatible reference-image workflows.
- Expanded the original documentation with per-node input and output information.
Troubleshooting comfyui-ysnodes
A Size Picker latent causes a dimension mismatch
- Check which generation model the workflow uses.
- In Size Picker or Size Picker Full, set
latent_typeto the matching option in the compatibility table above. - Check that the rest of the workflow is intended for the same model architecture, then run it again.
The author added latent_type in version 1.2.0 to address latent outputs that had previously always used the SD/SDXL format.
A connected image has an unexpected size
- Check whether
use_image_sizeis enabled in Size Picker or Size Picker Full. - Check
fit_megapixels: this determines whether the image keeps its exact dimensions or is resized toward the megapixel target. - Review the
megapixels,multiple,width_override, andheight_overridesettings before running the workflow again.
An image node does not provide the latent you need
- Check which scaling node you used.
- If you used ImageScaleToNormPixels, note that it outputs only
imageandmask. - Use ImageScaleToTotalPixels when you need its VAE-based latent outputs, and provide a suitable
vaefor encoding.
The notification sound does not play
- Check that
play_soundis enabled andsound_volumeis set appropriately. - If using a local sound, check that the file exists in
comfyui-ysnodes/web/assets/. - Make sure
sound_filenames that file correctly, or points to the browser-playable URL you intend to use. - Check that the extension’s notification frontend file is present in its
web/folder. The source documentation refers to the notification script by two different filenames, so check the filename in your copy of the extension rather than relying on one name.
Frequently asked questions
What is the difference between Size Picker and Size Picker Full?
Size Picker produces an empty latent for starting generation. Size Picker Full also provides a VAE-encoded latent_from_image output for workflows that begin with an image.
Which scaling node should I use?
Use ImageScaleToNormPixels when you need resized image and mask outputs only. Use ImageScaleToTotalPixels when you also need its latent outputs.
Does the extension include SD, Flux, or other generation models?
No. Its latent_type control helps prepare an appropriately shaped latent for a model you use elsewhere in ComfyUI.
Will Load Image Temporarily put my upload in /input?
No. The author describes its upload button as sending images to ComfyUI’s /temp folder instead.
Learn More about comfyui-ysnodes
- comfyui-ysnodes on GitHub provides the extension’s source and project information.
- ComfyUI on GitHub is a starting point for learning about the node-based application in which this extension runs.
Several nodes build on ideas from other open-source ComfyUI work. The original README credits Enashka, royceschultz, T8mars, chflame163, comfyanonymous, and MNeMiC Nodes.
