ComfyUI-Empty-Latent-Advanced-Plus

Empty Latent Advanced Plus for ComfyUI

An extended version of Empty Latent Advanced with 67 resolution presets, a category filter, model-aware empty latents and a compact UI.

Credits: Based on ComfyUI-Empty-Latent-Advanced by ameyukisora (copyright 2025), distributed under the MIT License. This is a community modification, not an official update from the original author. The original copyright and license are retained in LICENSE.

Features

• 67 presets grouped by SD 1.5, SDXL, HD 720p, FHD 1080p, 2K, QHD and 4K.

• Filter above Resolution: switch between categories without scrolling through the full list.

• Aspect ratios: the HD, FHD, 2K and 4K groups include 9:21, 9:16, 2:3, 3:4, 4:5, 1:1, 5:4, 4:3, 3:2, 16:9 and 21:9 (QHD has seven ratios).

• Model-aware Latent Type selector with three formats:

  | Latent Type | Channels | Spatial downscale | |---|---:|---:| | SD 1.5 / SDXL | 4 | 8× | | SD3 / Flux 1 / Qwen | 16 | 8× | | Flux 2 / Klein | 128 | 16× |

• Override mode for manual Width and Height; optional Ratio Lock.

• Selecting a resolution updates the manual Width / Height fields even when Override mode is active.

• Retains the original Latent, Width and Height outputs and the node identifier EmptyLatentAdvancedSelector.

• Preserves previous workflow widget storage order despite moving Filter visually above Resolution.

Installation

If the original Empty Latent Advanced is installed, close ComfyUI and back up its folder first. Do not install both editions at the same time: both register the same node type.

cd ComfyUI/custom_nodes
git clone https://github.com/dar06051990-oss/ComfyUI-Empty-Latent-Advanced-Plus.git

Alternatively, download this repository as ZIP and extract it inside ComfyUI/custom_nodes/.

Restart ComfyUI and hard-refresh your browser (Ctrl+F5) to reload the custom JavaScript.

The clone URL above will work after this repository has been created and published.

Usage

1. Add Latent → 📐 Empty Latent Advanced.
2. Pick Filter (e.g. 4K) and then Resolution.
3. Choose the correct Latent Type for the model used by your workflow.
4. Leave Mode as Use Preset, or set it to Override to enter custom dimensions.
5. Connect Latent to the appropriate latent input of your workflow; optionally connect Width and Height to nodes requiring image dimensions.

Important: Model type is selected manually; it is not auto-detected from the loaded checkpoint. This node only prepares an empty spatial latent: it does not convert models, add video/temporal latent support, or guarantee compatibility with arbitrary workflows.

Flux 2 and dimensions

Flux 2 uses a 16× spatial downscale. Requested dimensions that are not multiples of 16 are aligned to the nearest supported size; the Width/Height outputs report the actual generated dimensions. For example, 1920×1080 yields 1920×1088 with Flux 2.

Performance note

Generating at 2K/4K can require substantially more VRAM. The presence of a preset does not guarantee the selected model can generate that resolution on the available GPU.

Compatibility

• ComfyUI with PyTorch installed (no extra Python dependencies beyond the usual ComfyUI environment).
• The original node type, original preset labels and original outputs are retained.
• Saved workflows created with older versions are intended to remain compatible. If you keep both editions installed, duplicate node registration may cause conflicts.

Tests

Local smoke checks cover 67 presets across three latent formats, aspect-ratio override behavior, filter switching, and workflow value serialization. Run with an existing PyTorch environment and Node.js:

python tests/test_latent.py
node tests/test_ui.cjs

Credits and license

Original project: amenorira / ComfyUI-Empty-Latent-Advanced, by ameyukisora. Original copyright: © 2025 ameyukisora. Original MIT License preserved. See LICENSE.

Modified edition prepared for the dar06051990-oss GitHub account.