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

ModelSamplingFlux, ModelSamplingSD3 and ModelSamplingAuraFlow node in ComfyUI

The ModelSampling nodes set the noise schedule shift for flow-matching models (Flux, SD3.5, Wan, Hunyuan, Qwen-Image, Chroma, Lumina) so samplers spend the right amount of time on composition versus detail.
Node pack: ComfyUI core
Category: Utilities and workflow helpers

What ModelSamplingFlux, ModelSamplingSD3 and ModelSamplingAuraFlow does

Flow-matching models do not use the SD noise schedule; they have a timestep shift parameter that moves sampling effort towards the noisy (compositional) end. ModelSamplingFlux computes it from max_shift, base_shift and the latent size (the Flux Dev defaults 1.15 / 0.5). ModelSamplingSD3 takes a single shift value (3 for SD3.5, 5-8 for Wan and Hunyuan, 3-6 for Qwen-Image). ModelSamplingAuraFlow is the same for AuraFlow and Lumina-type models. Each sits between the model loader (or LoRA) and the sampler and outputs a patched MODEL.
Changing the shift is the most effective "secret" tuning knob for these models: a higher shift smooths and improves coherence of large shapes, a lower shift adds fine detail and texture (and sometimes artefacts). Distillation LoRAs and turbo models often need specific values quoted on their model pages.

Inputs

Name
Type
What it is
model
MODEL
From the loader or LoRA.
max_shift / base_shift (Flux)
FLOAT
1.15 and 0.5 default; raise max_shift for cleaner, lower for more detail.
width / height (Flux)
INT
Latent size used to compute the shift.
shift (SD3 / AuraFlow)
FLOAT
3 for SD3.5, 5-8 for Wan and Hunyuan.

Outputs

Name
Type
What it is
MODEL
MODEL
The model with the adjusted schedule, for 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 ModelSamplingFlux, ModelSamplingSD3 and ModelSamplingAuraFlow

  1. Insert the node between Load Diffusion Model (or Load LoRA) and KSampler.
  2. For Flux Dev keep the defaults and match width/height to your latent.
  3. For Wan set ModelSamplingSD3 shift 8 (480p) or 5 (720p).
  4. Render once with the default and once with a shift 2 higher to see the effect.
  5. Note the value in the Save Image prefix when comparing.

Settings and tips

  • Flux Schnell and Nunchaku/turbo variants usually work without the node.
  • Chroma and Flux Krea like max_shift around 1.0-1.3.
  • A higher shift helps at high resolutions, where small details otherwise dominate early steps.
  • If a LoRA page says "shift 3", this is the node where it goes.

Troubleshooting ModelSamplingFlux, ModelSamplingSD3 and ModelSamplingAuraFlow

Image is blurry or lacks texture

Why it happens
Shift too high for the resolution and step count.

How to fix it
Lower max_shift (Flux) or shift (SD3 node) by 1-2 and add steps.

Over-detailed, noisy, broken anatomy

Why it happens
Shift too low; the model spends too little time on composition.

How to fix it
Raise the shift; for Wan use 8 at 480p.

The node has no effect

Why it happens
It is not in the chain to the sampler, or it is placed before a LoRA that resets the patch (rare).

How to fix it
Place it last before the sampler and check the MODEL link.

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 ModelSamplingFlux, ModelSamplingSD3 and ModelSamplingAuraFlow

Do I need it for Flux?

Flux Dev works without it (defaults are applied internally), but the node lets you tune. Flux workflows from the ComfyUI examples include it.

Which node for Wan?

ModelSamplingSD3 with shift 5-8. Wan 2.2 graphs apply it to both the high and the low noise models.

Related nodes

Load Diffusion Model
ComfyUI core
Load Diffusion Model (UNETLoader) loads a bare denoising network such as Flux, SD3.5, Wan or HunyuanVideo from models/diffusion_models, with a weight_dtype option for fp8 to save VRAM.
Flux Guidance
ComfyUI core
Flux Guidance (FluxGuidance) writes the guidance value into the conditioning that Flux Dev was distilled to expect, replacing cfg; 3.5 is the default and 2 to 5 the useful range.
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.
WanImageToVideo and Wan 2.x video nodes
ComfyUI core
WanImageToVideo builds the conditioning and the empty video latent for a Wan 2.1/2.2 image-to-video render from a start image, a CLIP Vision embedding and the frame count; the surrounding graph uses Load Diffusion Model, Load CLIP, Load VAE and KSampler.
EmptySD3LatentImage
ComfyUI core
EmptySD3LatentImage creates the 16-channel empty latent that Flux, SD3 and SD3.5 need; it replaces Empty Latent Image in those workflows.
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.
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.
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.