EasyLoader (A1111)

If you came to ComfyUI from Automatic1111 or Forge and immediately missed having your checkpoint, VAE, LoRA, resolution and prompts all in one place, this is the node that gives it back. easy a1111Loader collapses the whole front end of a SD1.5/SDXL workflow - the noodle-salad of Load Checkpoint → Load VAE → Load LoRA → CLIP Text Encode ×2 → Empty Latent - into a single tidy box that emits one pipe wire.

It's the flagship "loader" of ComfyUI-Easy-Use, and the reason a lot of people install the pack at all. The pack's whole philosophy is fewer connections, and this is where you feel it most: one node in, one wire out, and the rest of your graph stays readable.

Why you'd reach for it

Because a bare-ComfyUI text-to-image setup has a lot of boilerplate that never changes, and dragging it out every time is tedious. This bundles it. There's also a genuine feature here beyond convenience: the a1111_prompt_style toggle reproduces how Stable Diffusion WebUI actually parses prompts - attention weighting, BREAK, the works - so an image you dialled in on A1111 comes out nearly identical in ComfyUI. That's not cosmetic; ComfyUI and A1111 weight prompts differently by default, and this closes the gap. (It needs ComfyUI_smZNodes installed to do it.)

How it works

Pick your models and type your prompts, and the node loads the checkpoint, applies the VAE and any LoRA, encodes both prompts, and builds an empty latent at your chosen size - then packs all of it into a single PIPE_LINE. That pipe is Easy-Use's connective tissue: it carries the model, CLIP, VAE, positive, negative and latent through one cable to the next stage (a presampling node, then easy kSampler). You wire loader → sampler and you're done.

The inputs that matter

Most fields are self-explanatory; the few worth calling out:

• ckpt_name - your SD1.5/SDXL checkpoint. The core choice.
• clip_skip - defaults to -2, which is the right default for most SDXL and anime-lineage checkpoints (the pack deliberately switched this from -1 a while back).
• resolution - a dropdown of sensible sizes, or set empty_latent_width/height yourself for custom.
• a1111_prompt_style - the toggle that matches WebUI prompt parsing. Leave it off for native ComfyUI behaviour; turn it on (with smZNodes) to reproduce A1111 results.

Optional optional_lora_stack and optional_controlnet_stack let you feed in a whole stack of LoRAs or ControlNets built elsewhere. Outputs are pipe (into your sampler), plus model and vae broken out for when you need them raw.

Installing it

ComfyUI Manager: search ComfyUI Easy Use, install, restart. Or cd ComfyUI/custom_nodes && git clone https://github.com/yolain/ComfyUI-Easy-Use, install requirements (install.bat / pip install -r requirements.txt), restart. For the a1111_prompt_style feature, also install shiimizu/ComfyUI_smZNodes.

Where people trip

The recurring one: turning on a1111_prompt_style without smZNodes installed - it won't behave and you'll wonder why your prompts don't match. Install smZNodes first. Second, this loader is built for SD1.5/SDXL; for Flux you want easy fluxLoader instead, because Flux's guidance model doesn't use a negative the way these do. And as with the whole pack: if the loader's widgets look scrambled after a ComfyUI frontend update, update Easy-Use - a lot of the changelog is exactly these frontend-compat fixes.