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

Upscale Latent and Upscale Latent By node in ComfyUI

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.
Node pack: ComfyUI core
Category: Latent

What Upscale Latent and Upscale Latent By does

Generating big images in one pass gives duplicated subjects, so ComfyUI workflows generate small, enlarge the latent, and sample again with a low denoise. Upscale Latent takes a LATENT and a target width and height; Upscale Latent By takes a scale factor instead. Both offer nearest-exact, bilinear, area, bicubic and bislerp interpolation.
Latent upscaling is cheap but blurry on its own; the second sampler pass at denoise 0.4-0.6 is what restores sharpness and adds detail. For the cleanest results many people decode, upscale in pixels with an ESRGAN model and re-encode instead; latent upscaling is faster and keeps more of the original look.

Inputs

Name
Type
What it is
samples
LATENT
The latent from the first sampler.
upscale_method
COMBO
nearest-exact, bilinear, area, bicubic or bislerp.
width / height (Upscale Latent)
INT
Target pixel size.
scale_by (Upscale Latent By)
FLOAT
Factor, for example 1.5 or 2.0.
crop
COMBO
disabled or center, when the aspect ratio changes.

Outputs

Name
Type
What it is
LATENT
LATENT
The enlarged latent for a second KSampler.
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 Latent and Upscale Latent By

  1. First KSampler at native size, denoise 1.0.
  2. Upscale Latent By, scale_by 1.5, method bislerp or bicubic.
  3. Second KSampler with the same model and prompts, denoise 0.45-0.55, 15-20 steps.
  4. VAE Decode the second latent.

Settings and tips

  • 1.5x is the sweet spot; 2x latent upscales need denoise 0.55+ to recover from blur.
  • bislerp and bicubic give the smoothest latents; nearest-exact keeps crisp edges for pixel art.
  • Use the same seed in both samplers for consistency, or a different one for extra variation.
  • Lower the second-pass cfg by 1 to avoid over-contrast.

Troubleshooting Upscale Latent and Upscale Latent By

The upscaled result is soft or blurry

Why it happens
No second sampler pass after the latent upscale, or denoise too low (under 0.35).

How to fix it
Add a KSampler after the upscale with denoise 0.45-0.6.

The second pass changes the image completely

Why it happens
denoise too high (over 0.7) or different prompts/cfg in the second sampler.

How to fix it
Lower denoise to 0.5 and keep prompts, model and cfg the same as the first pass.

Extra limbs or doubled details appear after upscaling

Why it happens
The second pass runs at a size far above the model training size with a high denoise.

How to fix it
Upscale 1.5x instead of 2x, lower denoise, or add a ControlNet Tile to pin the composition.

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 Latent and Upscale Latent By

Latent or pixel upscale?

Latent is faster and keeps the style; pixel (ESRGAN) plus re-encode gives sharper fine detail. Many workflows do pixel upscale then a light sampling pass.

Why does bislerp exist?

It interpolates on a sphere, which suits the normalised latent values better than linear blending.

Related nodes

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.
Upscale Image (using Model)
ComfyUI core
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.
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.
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.
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.
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.
Empty Latent Image
ComfyUI core
Empty Latent Image creates the blank latent canvas (width, height, batch_size) that text-to-image sampling starts from; dimensions must be multiples of 8 and match the model family.
Load Image
ComfyUI core
Load Image reads a picture from the ComfyUI input folder (or an upload) and outputs an IMAGE tensor plus a MASK taken from the alpha channel, the starting point for img2img, inpainting, ControlNet and IP-Adapter graphs.