Create Blank Frames

Sometimes you don't have a source video - you want to make one from scratch. Create Blank Frames generates a stack of solid-color frames at whatever size you want: the raw material for a slideshow, a title card sequence, a logo animation, or a test pattern you can hang the rest of your workflow off. It's one of the three nodes the pack's author actually documents in the README, which tells you it's considered a core building block rather than an afterthought.

The pack is ComfyUI Image to Video Inserter by Yeq6X. The workflow it's built around is: get a frame stack (from a video, or created blank), insert still images at specific frame positions with Multi Image Inserter, then re-encode. When you don't have a source clip, Create Blank Frames is the entry point - the "canvas" for the whole pipeline. The repo's example workflow wires exactly that: blank frames → assembler → inserter → video combine.

How it works

There's no mystery here. The node builds frame_count OpenCV frames of a solid color at width × height, then converts them to the ComfyUI IMAGE tensor format and hands them over. It's a generator, not a processor - nothing upstream, just pure output. If you want anything more than flat color on these frames, you composite it on later with an inserter or a blend node.

Inputs that matter

• width / height - frame dimensions, default 640×416, range 64–4096 in steps of 8. That 640×416 default (about a 1.54:1 frame) is a sensible low-res video canvas; bump it to match whatever your model or encoder expects.
• frame_count - how many frames, default 100, max 10,000. At 30 fps that's a touch over three seconds.
• color_r / color_g / color_b - the fill color as 0–255 RGB values, black by default. Want a pure black start for a video you'll add light overlays to? Leave it. Want a green-screen base for later chroma work? Dial it in.

The output is frames (IMAGE), a batch of identical solid-color frames ready for whatever's downstream.

Installing it

One install brings the whole pack:

cd ComfyUI/custom_nodes
git clone https://github.com/Yeq6X/ComfyUI-image-to-video-inserter

then restart ComfyUI - or just use ComfyUI Manager and search "Image to Video Inserter". There are no model weights to fetch and no API keys; the node only needs Pillow, numpy and OpenCV, which any working ComfyUI already has. If the pack doesn't load at all, the usual culprit is a missing opencv-python - pip install opencv-python in the ComfyUI environment fixes it.

When to reach for it

Reach for this when you're building a video timeline from nothing instead of editing one - placeholder frames to prototype a layout, a black base to drop title cards onto, or a solid test signal to check your encoder path before you spend a generation budget on real content. It won't win awards for what it does, but it's the honest, zero-drama way to start a frame stack that every video workflow in this pack can build on.