sm4ll Wrapper Lookbook Sampler - Paid API

Where the other two nodes in this pack do one garment per call, sm4ll Wrapper Lookbook Sampler (VTONLookbookNode) does the whole outfit. Feed it a single person photo and up to four garment images - shoes plus a dress plus accessories, cross-category is explicitly allowed - and the backend composes them into a grid layout and generates one try-on image of that person wearing everything at once. If you've ever wanted a "complete the look" page for an online store, this is that, as a node.

It's also the most expensive node in the pack: $0.15 per call at normal quality (3 credits), $0.30 at high (6 credits). That's three times the paid single-garment node, so the calculus only makes sense when the multi-garment composition is the actual deliverable. If you just need one shirt tried on, use VTONAPIPaidNode for a nickel; save the lookbook call for the full-outfit shot where manually compositing four separate try-ons in an editor would cost you an hour.

How it works

Same preprocessing family as the siblings - every image resized to 1.62 megapixels with Lanczos, forced to RGB - but the transport differs. Where the other nodes upload files to a session, the lookbook node base64-encodes the person and garment images and POSTs them directly to apiservice.yourmirror.io as one request. The backend lays out the garments, runs a specialized lookbook VTON model, and hands back the result either as a URL or a base64 data URI, which the node decodes into your IMAGE output. It's NOT_IDEMPOTENT like the rest of the pack, so every queue run is a fresh (billable) request.

Two things the README calls out that matter for results: the backend runs separate male and female pipelines based on the mode you pick, and unlike the single-garment nodes there's no mask input - no override for the automatic masking. If you're picky about masks, the lookbook node isn't where you want to be.

Inputs and outputs

The required set is small but the optional ones do the work:

• person_image - the model photo (required)
• garment_1 - first garment (required); garment_2, garment_3, garment_4 - add up to three more (optional). Optional garments under 100×100 pixels get silently skipped with a console warning, so check the log if your 3-garment call only shows two.
• mode - Male or Female (required). This isn't a suggestion box; it routes to a different processing pipeline.
• quality - Normal or High (required), and it directly doubles the price. Try Normal first.
• prompt - optional free-text guidance, empty by default. Worth using if you have a specific styling in mind.
• api_key - YourMirror key, ym_ + 29 characters, same one you use for the paid single-garment node. Same 1000-requests-per-hour limit.

Output is a single IMAGE tensor - Save Image or Preview, or downstream processing.

Install

Identical to the rest of the pack - installing once gets you all three nodes under sm4ll/VTON:

cd ComfyUI/custom_nodes/
git clone https://github.com/risunobushi/ComfyUI_sm4ll-Wrapper.git
cd ComfyUI_sm4ll-Wrapper
pip install -r requirements.txt

Restart ComfyUI. The requirements (torch, torchvision, numpy, Pillow, requests) are already present in stock ComfyUI, and there are no model downloads - you're buying compute per call, not disk space. ComfyUI Manager can install the pack if you search "sm4ll".

Troubleshooting

The pack's signature failure mode applies here harder than anywhere, because it costs money: a solid red 512×512 output means an error occurred, but ComfyUI treats it as a success. Read the console - the lookbook code is verbose by design, printing every garment it processes and every result it receives.

The rest is familiar territory: API key is required means the placeholder key is still in the field; Invalid API key format means it doesn't start with ym_ or isn't exactly 32 characters; Failed to upload is a network/backend availability problem. The result comes back as either a URL or base64, and the node handles both - if you see a "Failed to download result" in the console, retry the queue. And when a run "succeeds" with fewer garments than you expected, check the console for the skipped-garment warning before re-billing yourself.