Inputs Basic

AUNInputsBasic is the lighter sibling in the AUN Inputs family, and - read the pack's own notes - it's the one new workflows are supposed to use. Where the full AUNInputs node also tries to handle your save filenames, this one stops at the generation contract: load the checkpoint, apply a SpeedLoRA if you want, set your sampler settings, create the empty latent, done. No MainFolder, no Filename, no name mode outputs dragging the old naming system into your graph. The pack's guidance is that full Inputs nodes are "legacy directionally" and the Basic ones paired with AUN Save Image V2 are the cleaner, more future-proof way to build.

What you get

The outputs are the sampling essentials, nothing more:

• MODEL, CLIP, VAE - checkpoint split into its three parts for KSampler and text encoding.
• latent - the empty latent from your chosen aspect ratio or width/height.
• sampler, scheduler, cfg, steps, seed, width, height, batch size - as typed outputs you can route or display.
• ckpt name - the model name string for filenames.

That's a tidy replacement for Checkpoint Loader + Empty Latent + a pile of "set value" nodes, and it composes with a separate filename/save node (like AUN Save Image V2) instead of trying to be one.

Inputs: ckpt_name, sampler/scheduler/cfg/steps, aspect_ratio (presets or custom, with megapixels and multiple to compute resolution from a ratio), aspect_mode (Random/Swap/Original for dimension orientation), batch_size, seed, and the SpeedLoRA trio. clip_skip defaults to -1. The SpeedLoRA inputs (speed_lora, speed_lora_model, speed_lora_strength) are for the distilled fast-checkpoint LoRAs, which want a low CFG - hence the default cfg of 2 rather than the classic 7-ish.

Installing it

AUN ships all nodes in one pack:

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

Manual installs hitting ModuleNotFoundError: cv2 or piexif? pip install -r custom_nodes/AUN-ComfyUI-Nodes/requirements.txt.

Common issues

The biggest beginner stumble is the sheer number of widgets for what feels like a "basic" node - but every one maps to a KSampler setting you'd otherwise set elsewhere, so it's the same surface area, just in one place. Double-click the node body to collapse the output connections if it feels busy. And don't expect the full node's filename outputs here - if your workflow was built on MainFolder/Filename sockets from AUNInputs, this node won't drop-in replace it without rewiring your save side.