🌀 BV Regional Preview Send

The regional workflow's killer feature isn't drawing regions - it's the feedback loop. You draw a layout, render, look at what the model did with it, and adjust. The catch is that "look at what the model did" normally means alt-tabbing between the editor and the image preview and squinting. BV Regional Image Send (display name: BV Regional Preview Send) closes that loop by sending the rendered image back into the BV Regional Prompt editor as a background layer, while also previewing it and passing the image through. You see the render where you drew the regions.

How it works

Inputs: images (your rendered IMAGE tensor) and document_id (STRING). Outputs: images - the same tensor, passed through, so you can chain this node into the middle of a graph without breaking your image flow.

The document_id is the critical bit, and it's where people get stuck. The node sends the image to a specific regional document's editor - the editor for the document whose ID you provide. That's why the ID is required and why an empty value raises a clean error. You get the ID from the BV Regional Debug node's document_id output, or from the BV Regional Prompt node itself. Multiple documents, multiple editors, and each preview lands in the right one - that's the design. If you leave it blank, the node refuses rather than guessing, which is the right behavior for a feedback channel.

Wire it after your VAE decode (and after the compiler nodes, since it takes an IMAGE, not a latent). It previews the image in the normal ComfyUI image-output slot and ships it to the editor, so nothing about your usual review flow changes.

Why the background layer matters

The editor supports a background image with adjustable opacity - that's the surface your region geometry gets drawn over. Drop the render in as background and you can immediately see how your rectangles and brushes relate to what the model actually produced: did the subject grow past its region, did the interaction rectangle actually capture the contact point, is the empty center really empty? That visual comparison is the difference between tuning regional geometry by guesswork and tuning it by evidence.

Install

Search BV Node Pack in ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/BlackVortexAI/bv_nodepack.git

Restart, hard-refresh with Ctrl + F5. No dependencies.

The honest take

This is a small quality-of-life node, but it's the one that turns regional authoring from "draw blind, render, screenshot, squint" into a real iterate-on-geometry loop - arguably the best reason to adopt this pack over raw masked-conditioning workflows. The document_id requirement is the only friction; wire it from Debug once and it's set. If you want the image saved to disk as well as sent, that's the sibling BV Regional Image Save (Save Send) node.