RenderZero Video Start Frame Bridge

A RenderZero video workflow is an image-to-video workflow, which means something has to hand your video model its first frame. That's this node. RenderZero Video Start Frame Bridge loads the start frame the platform uploaded - the image your clip animates out of - and hands it to the model as a plain IMAGE tensor. Without it, your i2v workflow has nothing to start from, and your video model will (rightly) refuse to run.

How it works

The mechanism is deliberately minimal. It takes one string input, image_slot, resolves it against ComfyUI's input directory, EXIF-transposes the image, converts it to RGB, and returns it as a tensor. No resizing, no cropping, no canvas gymnastics - the frame arrives exactly as uploaded, and any scaling to your model's preferred input size happens downstream with normal ComfyUI image nodes. That's consistent with how the whole pack treats uploaded images: preserve fidelity at load, transform deliberately later.

The one opinionated bit is the error handling. If image_slot is blank, the node raises a clear "RenderZero video start frame is missing image_slot. Select or upload a start frame before queueing this workflow" error rather than failing silently with a confusing downstream crash. That's a small thing, and it's the right thing.

The input that matters

There's only one: image_slot (STRING). RenderZero fills it when you pick or upload a start frame in the platform UI; you won't normally type a filename here. The single output is image (IMAGE), which wires into your video model's start-image or input-image input - for Wan image-to-video, that's the conditioning input that seeds the clip.

Installing it

Same repo, same install as the rest of the pack. ComfyUI Manager - search "renderzero-comfyui-nodes" - Install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/jamesstothard/renderzero-comfyui-nodes renderzero_bridge_nodes

Restart ComfyUI. No requirements.txt, no model downloads - a few hundred lines of plain Python using only ComfyUI's bundled helpers.

Common issues

The most common failure is the most instructive: an empty image_slot means RenderZero hasn't uploaded a start frame, and the node will tell you so in plain language. Queue a start frame and it goes away. Second, because the frame is loaded at native size, a very large start image can balloon your memory usage before your model ever sees it - if you're feeding a 4K upload into a 480p video model, scale it down with ImageScale first. And remember the contract: every RenderZero video workflow requires exactly one of these, alongside the Video Prompt Bridge and Video Output, and the image-side bridge nodes must not appear in the same graph.