Nodes/ComfyUI Easy Use/EasyLoader (Flux)
ComfyUI Node Runs on cloud

EasyLoader (Flux)

The whole front end of a Flux workflow in one node

By yolain·Created 3 years ago·Updated 8 days ago· 2,633
EasyLoader (Flux)
  • model_override
  • clip_override
  • vae_override
  • optional_lora_stack
  • optional_controlnet_stack
  • pipe
  • model
  • vae
ckpt_name
vae_name
lora_name
lora_model_strength1.00
lora_clip_strength1.00
resolution1024 x 1024
empty_latent_width1024
empty_latent_height1024
positive
batch_size1

Flux is the model most people reach for now, and setting it up in bare ComfyUI is a small chore: checkpoint (or the split UNet/CLIP/VAE), the dual text encoders, a LoRA loader, an empty latent sized right. easy fluxLoader folds all of that into one node that spits out a single pipe. If you liked how easy a1111Loader cleaned up SD1.5/SDXL setups, this is the same idea aimed at Flux.

It's part of ComfyUI-Easy-Use's loaders family, whose entire selling point is fewer wires and less boilerplate. For Flux specifically that matters, because Flux workflows tend to sprawl.

Why you'd reach for it

Convenience, mostly, and the right convenience. One node holds your Flux checkpoint, VAE, a LoRA with its strengths, your resolution, your positive prompt and batch size - then hands the next stage everything it needs on one cable. Notice what's not there: a negative prompt field. That's correct, not an oversight - Flux.1-dev is a guidance-distilled model that doesn't take a real negative the way SD does, so the loader doesn't pretend it does. If you've been dragging a negative around your Flux graphs out of habit, this is the node quietly telling you to stop.

How it works

You choose your models and prompt; the node loads and assembles them into a PIPE_LINE - the Easy-Use bundle carrying model, CLIP, VAE, conditioning and latent through a single wire. From there it goes into a presampling node and then easy kSampler, same as the other loaders. The optional_lora_stack and optional_controlnet_stack inputs let you attach stacks built elsewhere, and the three override inputs (model_override, clip_override, vae_override) let you feed components in separately when you're loading Flux from split files rather than a single checkpoint - which, given how Flux is usually distributed, you often are.

The inputs that matter

  • ckpt_name - your Flux checkpoint (or use the overrides for split UNet/CLIP/VAE).
  • lora_name + lora_model_strength / lora_clip_strength - one LoRA inline; stack more via optional_lora_stack.
  • resolution - dropdown of good sizes, default 1024 x 1024 (Flux likes megapixel-ish); or set empty_latent_width/height.
  • positive - your prompt. No negative, by design.

Outputs: pipe into the sampler, plus model and vae broken out.

Installing it

ComfyUI Manager: search ComfyUI Easy Use, install, restart. Or manually: cd ComfyUI/custom_nodes && git clone https://github.com/yolain/ComfyUI-Easy-Use, install requirements (install.bat / pip install -r requirements.txt), restart. The Flux model files themselves are a separate download - the checkpoint, the two text encoders and the VAE - and they're large; this node loads them, it doesn't fetch them.

Where people trip

The most common confusion is the missing negative prompt - people go looking for it and assume the node is broken. It isn't; Flux just works that way. Second, if you're loading Flux from separate UNet/CLIP/VAE files (the common case for fp8/GGUF setups), use the override inputs rather than expecting ckpt_name to find a single all-in-one checkpoint. Third, the usual pack caveat: a frontend-version bump can scramble Easy-Use widgets, so if things look off after updating ComfyUI, update the pack too.

CategoryEasyUse/Loaders

Inputs (15)

NameTypeDefaultDescription
ckpt_nameCOMBO1 options: None
vae_nameCOMBO1 options: Baked VAE
lora_nameCOMBO1 options: None
lora_model_strengthFLOAT1.00-10–10
lora_clip_strengthFLOAT1.00-10–10
resolutionCOMBO1024 x 102431 options: width x height (custom), 512 x 512, 512 x 768, 576 x 1024, 768 x 512, 768 x 768, +25
empty_latent_widthINT102464–16384
empty_latent_heightINT102464–16384
positiveSTRING
batch_sizeINT11–64
model_overrideoptMODEL
clip_overrideoptCLIP
vae_overrideoptVAE
optional_lora_stackoptLORA_STACK
optional_controlnet_stackoptCONTROL_NET_STACK

Outputs (3)

NameTypeDescription
pipePIPE_LINE
modelMODEL
vaeVAE