Image RGB (QQ)

Image RGB (QQ) does one thing: it makes a solid-color image. No model, no sampler, no download - you give it a width, a height, and an R,G,B color, and out pops an IMAGE tensor filled with that exact color. ComfyUI's built-in Empty Image already does black, so this node's entire reason to exist is that any color is available and it comes with a batch size. Sounds trivial until you've wired a workflow where "please generate a 480x832 mid-gray frame" was previously a multi-node detour.

Why you'd actually reach for it

This pack is built around Wan video workflows, and that's where a color node earns its keep. A flat frame is the honest placeholder for I2V when you're testing the pipeline but haven't picked a first frame yet - or you want to composite something over a clean background and need a constant base layer. Convert the IMAGE to a mask and you have a uniform matte to drive inpainting or VACE regions. Even simpler: it's the cheapest way to sanity-check that your resize-and-encode chain handles a given resolution correctly, since the whole batch is generated in a few milliseconds and costs nothing but a slice of VRAM.

It also plugs into the pack's own tooling - the sibling nodes do color matching and reference prep, and a known-color frame is the input those expect. Feed the IMAGE into any encoder, mask-from-color node, or sampler's conditioning path and treat it like any other image.

How it works

The mechanism is dead simple, and worth knowing because it explains the only two ways you can trip it. The node parses your color_rgb string by splitting on commas, divides each channel by 255, and fills a [batch, height, width, 3] tensor with those values. ComfyUI images live as float tensors in 0–1 range, so "128,128,128" becomes 0.502 gray. It builds the tensor on ComfyUI's intermediate device with the appropriate intermediate dtype, so it doesn't churn your GPU unnecessarily.

The inputs that matter

• width / height - output resolution, 1 to 8192. Default 512 each.
• batch_size - 1 to 4096. More frames, more VRAM, linear in the math.
• color_rgb - a string like "128,128,128". This is the whole point.
• Output: IMAGE - a float tensor in 0–1 range, ready to wire anywhere an image is accepted.

Installing it

The pack is on ComfyUI Manager - search "ComfyUI-SA-Nodes-QQ" - or clone it:

cd ComfyUI/custom_nodes
git clone https://github.com/siraxe/ComfyUI-SA-Nodes-QQ

Restart ComfyUI. requirements.txt is empty, so there's nothing to pip install and no model files. One trap: this pack was renamed from ComfyUI-WanVideoWrapper_QQ (and its folder was wanwrapper_qq). If you had the old install, delete that folder first - stale workflows can silently pull the old package and give you duplicate node errors. The README's own install section still shows the old clone URL; use the one above.

Common issues

• "color_rgb must be in format 'R,G,B'" - you typed a hex color or a single value. It needs comma-separated integers.
• Values outside 0–255 - the code divides by 255 and does not clamp, so 300 produces a channel above 1.0. Keep channels in range.
• It looks black / white and you expected otherwise - check the string, not the node. "0,0,0" is black, "255,255,255" is white, and the default is mid-gray.

That's the whole node. It's the kind of utility you don't think about until you need a specific color frame at a specific resolution, at which point it's exactly the tool you want.