Faishme Mannequin to Model Loader

Calling this a "Model Loader" is generous. It loads no model weights, calls no API, and touches no file beyond reading the filename of a LoRA you already have installed. What Faishme Mannequin to Model Loader actually does is write your prompt for you: pick a garment or product LoRA, pick a pose, and it hands back the model name and a ready-to-use pose prompt. That's the whole job, and honestly it's a useful little job.

Where it fits

It's part of the Faishme fashion workflow - the same product-pipeline-turned-public pack that gives you Load Fashion Model. The idea: you've trained a LoRA on a specific garment or mannequin, and you want to generate images of it from different angles. Rather than hand-typing "full body portrait of spring_dress_01 in side view" for every combination, this node builds that string from a dropdown. It turns the tedious part of batch product photography into a configurable graph.

How it works

lora_name is a dropdown of every LoRA in your ComfyUI/models/loras folder. The node strips the extension to get a bare model name, then looks up your chosen pose_hint in a small prompt dictionary and produces something like full body portrait of spring_dress_01 in side view. That's the entire mechanism - three string outputs, no inference:

• model_name - the LoRA filename without its extension. Handy as a trigger word if your LoRA was trained with one.
• pose_prompt - the assembled pose sentence, meant to go straight into a CLIP Text Encode positive prompt.
• pose_hint - the raw hint you picked (full-front, upper-side, closeup, and so on), passed through so you can drive other nodes' logic with it.

The pose_hint dropdown has seven options: full-front, full-side, full-back, upper-front, upper-side, upper-back, and closeup.

The input that matters

Really, just one: pose_hint. The lora_name dropdown is mostly there to name the subject. The output you'll actually wire is pose_prompt → CLIP Text Encode → your sampler's positive conditioning. If you want a batch of angles, feed the same LoRA into several of these nodes with different pose_hint values, or rerun with a different hint - that's exactly the workflow it was built for.

Installing it

Same as the rest of the pack: search "ComfyUI_faishme" in ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/AkashKarnatak/ComfyUI_faishme

Restart ComfyUI and it shows up under the FaishmeNodes category. Note you still load the LoRA yourself with your normal LoRA loader - this node just names it in a prompt. Also note the pack is brand-new and has essentially zero community footprint; this is the author's internal fashion tooling made public, so expect it to match their pipeline's conventions rather than yours.

Gotchas

Because the node is a prompt builder and not a real loader, it can't check that the LoRA it names actually contains the garment you think it does - you're trusting your own training. And the pose prompts assume a fashion-photography framing; for a general character LoRA, "full body portrait of X in front view" is fine, but the phrasing is tuned for catalog-style output. Keep pose_hint in mind as the knob that changes the whole look of the output - it's the only one that matters.