Inputs

AUNInputs is the node the whole pack is named around: one node that loads your checkpoint, applies a SpeedLoRA, creates the empty latent, sets sampler/scheduler/cfg/steps, picks an aspect ratio, and prepares your save filename - and then spills all of it out as sockets for the rest of the graph. It's an "all-in-one" in the strongest sense: in the stock ComfyUI setup, everything this node does is five or six nodes. Here it's one node with a wall of widgets.

That's the pitch, and it's a real one for people who build tidy, compact workflows - which is exactly the audience AUN targets. But be aware of the catch before you commit: the pack itself now calls the full Inputs family "legacy directionally." The recommended path for new workflows is the Basic input nodes plus AUN Save Image V2, because the full node's save-prep outputs (MainFolder, Filename, prefix, date format, name mode) are the older naming contract. This node still works and is kept compatible, but it's a "great while it lasts" situation, not a "build your future on this" one.

What it outputs (and why that matters)

The outputs are the interesting part, because they replace whole chains:

• MODEL, CLIP, VAE - the loaded checkpoint, split out ready for a KSampler and CLIP Text Encode.
• latent - the empty latent, sized from your aspect_ratio / width / height.
• sampler, scheduler, cfg, steps, seed, width, height, batch size - all the sampling parameters as typed outputs, so you can route them anywhere or display them.
• ckpt name - the model name as a string, for filenames.
• The save-prep outputs - MainFolder, Filename, prefix, date format, name mode - for feeding an AUN filename builder or saver.

Inputs worth knowing: ckpt_name, sampler/scheduler/cfg/steps, aspect_ratio (presets or custom), seed, and the SpeedLoRA trio (speed_lora, speed_lora_model, speed_lora_strength) - SpeedLoRA is a real thing, and its proper CFG is low, so don't be surprised the default cfg is 2. clip_skip defaults to -1.

Installing it

Part of the AUN pack, one install for everything:

• ComfyUI Manager: search "AUN ComfyUI Nodes", install, restart.
• Manual: cd custom_nodes && git clone https://github.com/loz2754/AUN-ComfyUI-Nodes, then restart.

Deps (piexif, opencv-python-headless, imageio-ffmpeg, requests) come with Manager; manually, pip install -r custom_nodes/AUN-ComfyUI-Nodes/requirements.txt if you see ModuleNotFoundError: cv2.

Common issues

Two things get people. First, the widget count is overwhelming - that's by design, and double-clicking the node body collapses the connection display ("collapse connections" mode) so the graph stays readable. Second, if you migrate an old workflow that used AUNInputsRefine and the SpeedLoRA widgets look wrong after loading, the README warns the input set changed and you may need to re-check or reconnect SpeedLoRA widgets - worth knowing before you panic at a changed layout.