Model
Trends
.ai
Model
Trends
.ai
the best open source ai models

Upscale Image (using Model) node in ComfyUI

Upscale Image (using Model) runs an ESRGAN-style upscaler (4x-UltraSharp, RealESRGAN, 4x_NMKD-Siax, 4x_foolhardy_Remacri) over an IMAGE and returns it enlarged by the model's fixed factor with added detail.
Node pack: ComfyUI core
Category: Upscaling

What Upscale Image (using Model) does

This node applies a trained super-resolution network loaded by Load Upscale Model. The factor is fixed by the model (most are 4x, some 2x or 1x). The picture is processed in tiles so even large images fit in VRAM, then stitched. The result is sharper and more detailed than interpolation but cannot change the content; it is a pixel-space operation with no prompt.
A typical finishing chain is VAE Decode -> Upscale Image (using Model) with 4x-UltraSharp -> Upscale Image By 0.5 (back to 2x) -> Save Image. For more realism you can send the enlarged picture through VAE Encode and a low-denoise KSampler (0.2-0.35), which is what Ultimate SD Upscale automates.

Inputs

Name
Type
What it is
upscale_model
UPSCALE_MODEL
From Load Upscale Model.
image
IMAGE
The picture to enlarge.

Outputs

Name
Type
What it is
IMAGE
IMAGE
The enlarged picture (source size times the model factor).
AD
Skip the setup: ComfyUI workflows preinstalled on the cloud
BitVector runs ready-made ComfyUI workflows on its own GPUs. No install, no missing nodes, no red boxes. Open it on your phone or laptop and generate in a minute.

How to use Upscale Image (using Model)

  1. Put a .pth or .safetensors upscaler into models/upscale_models.
  2. Add Load Upscale Model and pick it.
  3. Add Upscale Image (using Model); connect the model and the IMAGE from VAE Decode.
  4. Follow with Upscale Image By 0.5 if 4x is more than you need.
  5. Save, or VAE Encode and run a low-denoise pass for extra realism.

Settings and tips

  • 4x-UltraSharp is the all-rounder; 4x_NMKD-Siax for photos; 4x-AnimeSharp for anime; RealESRGAN_x4plus_anime_6B is fast.
  • A 1x model such as 1x-ITF-SkinDiffDetail is a sharpening pass without resizing.
  • Upscaling before the face detailer gives it more pixels to work with.
  • The node tiles automatically; if it still runs out of memory, downscale the input first.
  • ESRGAN models cannot fix blur from a bad render; fix the render first.

Troubleshooting Upscale Image (using Model)

torch.OutOfMemoryError during the upscale

Why it happens
The tile size falls back to large values for huge inputs, or other models are still loaded in VRAM.

How to fix it
Downscale the input, run the upscale in a separate queue after unloading models (Free memory in the menu), or start with --lowvram.

Error: upscale model is not a supported architecture

Why it happens
The file is not an ESRGAN/RRDB-type model (for example a checkpoint or a SwinIR variant the loader does not know), or it is a broken download.

How to fix it
Download a model from the OpenModelDB in .pth or .safetensors format; ComfyUI supports ESRGAN, RealESRGAN, SwinIR, HAT, DAT, SPAN and many others through spandrel. Update ComfyUI for new architectures.

Result looks oversharpened or plastic

Why it happens
Many 4x models exaggerate edges, especially on already clean renders.

How to fix it
Try a softer model (4x_foolhardy_Remacri, RealESRGAN_x4plus) or blend the result with a bicubic upscale using Image Blend at 0.5.

Visible grid lines in the output

Why it happens
Tile seams; rare in core but visible with certain models at low overlap.

How to fix it
Update ComfyUI; or use Ultimate SD Upscale with seam fix for a diffusion-based pass.

AD
Run this workflow from your phone
Every workflow on this page is preinstalled on BitVector with the models and custom nodes already in place. Pick one, type a prompt, done. Zero setup, nothing to download.

Questions about Upscale Image (using Model)

Can I choose a 2x factor?

The factor is part of the model. Use a 2x model, or a 4x model followed by Upscale Image By 0.5.

Where do I find models?

OpenModelDB lists hundreds with sample comparisons. Put them in models/upscale_models.

Is this the same as hires fix?

No. Hires fix re-samples in latent space with the prompt; this node only enlarges pixels. Many workflows use both.

Related nodes

Load Upscale Model
ComfyUI core
Load Upscale Model (UpscaleModelLoader) loads an ESRGAN-style upscaler such as 4x-UltraSharp or RealESRGAN from models/upscale_models for use with Upscale Image (using Model).
Upscale Image and Upscale Image By
ComfyUI core
Upscale Image (ImageScale) and Upscale Image By resize an IMAGE with classic interpolation (nearest, bilinear, bicubic, lanczos, area) to an exact size or by a factor, with no AI model involved.
Ultimate SD Upscale
ComfyUI_UltimateSDUpscale
Ultimate SD Upscale enlarges a picture with an ESRGAN model, then re-samples it tile by tile through the diffusion model with your prompt at low denoise, adding real detail at 2x-4x without running out of VRAM.
Upscale Latent and Upscale Latent By
ComfyUI core
Upscale Latent resizes a latent to a new size (or by a factor) so a second KSampler can add detail at higher resolution: the classic hires fix.
AD
Tired of fixing nodes? Let the cloud do it
BitVector keeps hundreds of ComfyUI workflows installed, updated and tested on fast cloud GPUs. No Python, no CUDA errors, no VRAM limits. Works from any browser.

More ComfyUI nodes

CLIP Text Encode (Prompt)
ComfyUI core
CLIP Text Encode turns a text prompt into CONDITIONING using the model text encoder. One node holds the positive prompt, a second one the negative prompt.
ComfyUI Manager
ComfyUI-Manager
ComfyUI Manager is the extension that installs, updates and fixes custom node packs and models from inside the interface, resolves missing nodes in imported workflows and snapshots your setup.
KSampler
ComfyUI core
KSampler runs the denoising loop: it takes the model, prompts and a latent and produces the finished latent image, controlled by seed, steps, cfg, sampler, scheduler and denoise.
Load Checkpoint
ComfyUI core
Load Checkpoint (CheckpointLoaderSimple) opens a .safetensors or .ckpt model file and hands out the three parts every workflow needs: the diffusion MODEL, the CLIP text encoder and the VAE.
Save Image
ComfyUI core
Save Image writes the IMAGE tensor to ComfyUI/output as a PNG, with the whole workflow embedded in the file metadata so the picture can be dragged back into ComfyUI to restore the graph.
VAE Decode
ComfyUI core
VAE Decode converts the sampled LATENT into a pixel IMAGE with the VAE; VAE Decode (Tiled) does the same in tiles for very large images.