<img src="https://img.shields.io/badge/version-1.2.0-blue.svg" alt="Version 1.2.0"> <img src="https://img.shields.io/badge/type-skill-8A2BE2.svg" alt="Skill"> <img src="https://img.shields.io/badge/dependencies-none-brightgreen.svg" alt="No Dependencies"> <img src="https://img.shields.io/badge/nodes-9-orange.svg" alt="Nodes"> <img src="https://img.shields.io/badge/license-MIT-green.svg" alt="MIT License"> <h1 align="center">comfyui-ysnodes</h1> A lightweight ComfyUI custom node pack — every node is self-contained with zero dependency on external custom node packs.

Overview • Features • Installation • Quick Start • Nodes • FAQ • Contributing • Changelog • Credits • License

📖 Overview

comfyui-ysnodes is a personal pack of custom ComfyUI nodes covering resolution picking, sigma scheduling, reference conditioning, image scaling, and workflow utilities. Every node is a self-contained Python class — no external custom node packs required.

Who it's for: ComfyUI users who want a clean, dependency-free toolkit for common workflow tasks.

✨ Features

Zero external dependencies — every node is a self-contained Python class. No nhknodes, no extra installs
Multi-model latent support — latent_type selector for SD/SDXL (4ch), SD3/AuraFlow (16ch), Flux (16ch), and Flux2 (128ch)
Dynamic resolution picking — 17 aspect ratios × megapixel target × rounding multiple (replaces 35 fixed presets)
Browser notifications — workflow completion alerts with sound + OS notification
Temp folder image loading — load throwaway reference images from /temp with a built-in upload button

Node Index

| # | Node | Category | Description | |---|---|---|---| | 1 | ⚡ Sigma Generator | sampling | KSamplerSelect + BasicScheduler + SplitSigmas in one node | | 2 | 📐 Size Picker | utility | Dynamic aspect-ratio resolution picker with megapixel-based sizing | | 3 | 📐 Size Picker Full | utility | Size Picker with VAE encoding — outputs empty latent and image latent | | 4 | 🔗 Reference Conditioning | conditioning | Attaches a reference image latent to conditioning (Flux Kontext-style) | | 5 | 🔁 Redraw Size Calculator | image | Auto-scale image and mask to a clamped megapixel range | | 6 | 🔔 Purge & Notification | utility | Purge VRAM cache and send browser sound + OS notification | | 7 | 🖼 ImageScaleToTotalPixels | image | Scale image + mask to target MP with VAE encoding and passthrough switch | | 8 | 🖼 ImageScaleToNormPixels | image | Scale image + mask with scale_by + megapixels (no VAE encoding) | | 9 | 📂 Load Image Temporarily | image | Load images from ComfyUI's /temp folder with upload button |

📦 Installation

From GitHub

cd ComfyUI/custom_nodes
git clone https://github.com/MrYassinox/comfyui-ysnodes.git

Manual

Download and extract the comfyui-ysnodes/ folder
Place it in ComfyUI/custom_nodes/
Restart ComfyUI — all nodes appear under YSNodes/ in the node menu

[!NOTE] No additional dependencies. The pack uses only Python stdlib, PyTorch, and ComfyUI built-ins.

📁 Project Structure

ComfyUI/
└── custom_nodes/
    └── comfyui-ysnodes/
        ├── __init__.py                    # Package entry, auto-discovers nodes/, version="1.1.0"
        ├── README.md                      # This file
        ├── LICENSE                        # MIT License
        ├── requirements.txt               # Empty (stdlib + comfy + torch only)
        ├── web/
        │   ├── purge_notification_sg.js   # Frontend: sound + OS notification via WebSocket
        │   ├── load_image_temporarily_sg.js  # Frontend: upload button for /temp images
        │   └── assets/
        │       ├── notify.mp3             ← add your own sound file here
        │       └── notify.wav
        └── nodes/
            ├── __init__.py                # Auto-exports NODE_CLASS_MAPPINGS from all *.py
            ├── sigma_generator.py
            ├── size_picker.py
            ├── size_picker_full.py
            ├── redraw_size.py
            ├── purge_notification.py
            ├── image_scale_total_pixels.py
            ├── image_scale_to_norm_pixels.py
            ├── reference_conditioning.py
            └── load_image_temporarily.py

🧩 Nodes

⚡ Sigma Generator

YSNodes/sampling

KSamplerSelect + BasicScheduler + SplitSigmas in one node. Connect directly to SamplerCustom.

| Input | Output | |-------|--------| | model, sampler_name, scheduler, steps, skip_start_step, denoise | SIGMAS, SAMPLER |

📐 Size Picker

YSNodes/utility

Dynamic aspect-ratio resolution picker. Two modes: computed (aspect_ratio + megapixels) or image-driven (use_image_size=true). Supports fit_megapixels to resize the image to the megapixel target while preserving its aspect ratio.

| Input | Output | |-------|--------| | use_image_size, fit_megapixels, aspect_ratio, megapixels, multiple, width_override, height_override, batch_size, latent_type, image (opt) | image, width, height, batch, latent |

latent_type: SD / SDXL (4ch ÷8) · SD3 / AuraFlow (16ch ÷8) · Flux (16ch ÷8) · Flux2 (128ch ÷16)

<details> <summary>17 aspect ratios</summary>

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 · 2:1

</details>

📐 Size Picker Full

YSNodes/utility

Size Picker + VAE encoding. Outputs two latents: latent_empty (txt2img) and latent_from_image (VAE-encoded, for img2img/inpaint).

| Input | Output | |-------|--------| | Same as Size Picker + vae (opt) | image, width, height, batch, latent_empty, latent_from_image |

🔗 Reference Conditioning

YSNodes/conditioning

Attaches a reference image latent to conditioning — for Flux Kontext-style workflows. Optional megapixel scaling before encoding.

| Input | Output | |-------|--------| | conditioning, pixels, vae, upscale_method, megapixels, multiple, enable_megapixels, mask (opt) | conditioning, image, mask, latent |

🔁 Redraw Size Calculator

YSNodes/image

Scales an image to fit within a clamped megapixel range (14 presets from SD1.5 to 4K UHD). Outputs both resized and original versions.

| Input | Output | |-------|--------| | image, upscale_method, min_preset, max_preset, min_override, max_override, mask (opt) | image_resized, mask_resized, image_original, mask_original, width_resized, height_resized, width_original, height_original, target_megapixels |

🔔 Purge & Notification

YSNodes/utility

End-of-workflow node: purge VRAM cache + browser sound + OS notification. Place after your last Save Image node.

| Input | Output | |-------|--------| | purge_cache, purge_models, alert_mode, system_notification, notification_text, play_sound, sound_volume, sound_file, signal (opt) | signal |

[!TIP] Place your sound file in comfyui-ysnodes/web/assets/. Supports .mp3, .wav, or any browser-playable URL.

🖼 ImageScaleToTotalPixels

YSNodes/image

Scale image + mask to target megapixels with VAE encoding. Outputs both pixel and latent versions.

| Input | Output | |-------|--------| | pixels, upscale_method, megapixels, enable_megapixels, mask (opt), vae (opt) | image, latent_image, mask, latent_mask |

🖼 ImageScaleToNormPixels

YSNodes/image

Two-step scaling (scale_by + megapixels) without VAE encoding. scale_by=1.0 + enable_megapixels=false = passthrough.

| Input | Output | |-------|--------| | pixels, upscale_method, scale_by, megapixels, multiple, enable_megapixels, mask (opt) | image, mask |

[!NOTE] TotalPixels vs NormPixels: TotalPixels includes VAE encoding (outputs latents). NormPixels skips VAE — use it when you only need resized image/mask.

📂 Load Image Temporarily

YSNodes/image

Loads images from ComfyUI's /temp folder. Companion JS adds an upload button — files go directly to /temp, never cluttering /input.

| Input | Output | |-------|--------| | image (dropdown + upload) | image, mask, width, height |

❓ FAQ

<details> <summary>Why don't I hear the notification sound?</summary>

Two things must be in place:

web/purge_notification.js must be present in the comfyui-ysnodes/ folder — ComfyUI auto-serves it
A sound file (e.g. notify.mp3) must exist in comfyui-ysnodes/web/assets/

If using a custom filename or URL, update the sound_file input accordingly.

</details> <details> <summary>What is the difference between Size Picker and Size Picker Full?</summary>

Size Picker → outputs an empty latent only. Use for txt2img
Size Picker Full → outputs two latents: latent_empty (txt2img) and latent_from_image (VAE-encoded image for img2img / inpainting)

Size Picker Full has one extra optional input: vae. All other inputs are identical.

</details> <details> <summary>Do I need to edit `__init__.py` when adding a new node?</summary>

No. The main __init__.py auto-discovers every .py file inside nodes/. Just drop a new file there, restart ComfyUI, and your node appears automatically.

</details> <details> <summary>What is the difference between ImageScaleToTotalPixels and ImageScaleToNormPixels?</summary>

ImageScaleToTotalPixels → scales to a target megapixel count with VAE encoding — outputs image, latent_image, mask, latent_mask
ImageScaleToNormPixels → scales with scale_by + megapixels without VAE encoding — outputs image, mask only

Use TotalPixels when you need latents for sampling. Use NormPixels when you just need a resized image/mask (e.g. for conditioning inputs, previews, or chaining into other nodes).

</details> <details> <summary>Which latent_type should I use?</summary>

| Model | latent_type | |-------|-------------| | SD 1.5, SD 2.x, SDXL | SD / SDXL | | SD3, AuraFlow | SD3 / AuraFlow | | Flux (standard) | Flux | | Flux 2 | Flux2 |

</details>

🤝 Contributing

Create a new .py file in nodes/
Define your node class + register at the bottom:

NODE_CLASS_MAPPINGS = {"MyNewNode": MyNewNode}
NODE_DISPLAY_NAME_MAPPINGS = {"MyNewNode": "✨ My New Node"}

Restart ComfyUI — done ✅

📝 Changelog

v1.2.0

NEW: ImageScaleToTotalPixels — scale image + mask to target MP with VAE encoding and passthrough switch
NEW: ImageScaleToNormPixels — scale image + mask with scale_by + megapixels (no VAE encoding)
NEW: Load Image Temporarily — load images from ComfyUI's /temp folder with upload button
Added: latent_type input to Size Picker / Size Picker Full — selects latent format per model architecture (SD/SDXL, SD3/AuraFlow, Flux, Flux2)
Added: fit_megapixels input to Size Picker / Size Picker Full — controls whether connected image keeps exact size or gets resized to megapixel target
Added: image output to Size Picker / Size Picker Full — passthrough of connected image, or blank placeholder at computed size
Fixed: Size Picker / Size Picker Full — latent output was always SD/SDXL format regardless of target model, causing dimension mismatches with Flux2 and other non-SD architectures

v1.1.0

BREAKING: Size Picker — replaced 35 fixed resolution presets with dynamic aspect-ratio + megapixel + multiple system (17 aspect ratios)
BREAKING: Size Picker Full — same dynamic sizing as Size Picker, with latent_empty and latent_from_image outputs
NEW: Reference Conditioning — attaches reference image latent to conditioning (Flux Kontext-style)
Updated: README fully rewritten with per-node I/O tables, presets, and credits

🙏 Credits

Many of these nodes are extracted, merged, and improved from existing open-source ComfyUI nodes and subgraphs. This pack would not exist without the work of the original authors.

| Node | Based on | Author | |------|----------|--------| | Size Picker | 📐 Size Picker | Enashka | | Purge & Notification | Unified Notification + PurgeVRAM + LayerUtility | royceschultz · T8mars · chflame163 | | ImageScaleToTotalPixels | ImageScaleToTotalPixels | comfyanonymous | | Redraw Size Calculator | ImageScaleToTotalPixels + subgraph | comfyanonymous | | Sigma Generator | KSamplerSelect + BasicScheduler + SplitSigmas | comfyanonymous | | Reference Conditioning | VAEEncode + ReferenceLatent | comfyanonymous | | ImageScaleToNormPixels | ImageScaleToNormPixels + ImageScaleToTotalPixels | comfyanonymous | | Load Image Temporarily | Load Image Temporarily | MNeMiC Nodes |

All nodes rewritten as self-contained Python classes with zero external dependencies.

📄 License

⭐ Star on GitHub if you find this useful