Storyboard · Frame Grid

You've got a ten-shot sequence in your head and no desire to build ten CLIP Text Encode nodes, ten EmptyLatentImage nodes, and a mental spreadsheet tracking which prompt goes with which resolution. That's the job FrameGrid exists for. It's the pre-production node: you type your shot list once - name, prompt, aspect ratio per frame - and it hands you the active frame's prompt plus the width/height integers you need to generate it.

It's the flagship of Storyboard Suite (repo: Smyshnikof/comfyui-storyboard-suite), a small, dependency-free pack of "planning" nodes from @itsmyshnikov. No GPU, no model, no API key. It's a glorified spreadsheet wearing a ComfyUI skin, and that's the whole point: the pack's framing is that deciding what you're going to generate before you generate it should happen inside the graph, not in a separate doc you'll lose.

How it works

The frames_data widget holds a JSON array - each frame is an object with an id, a name, a prompt, and an aspect. A select value picks the active frame, and the node computes its resolution from the aspect ratio: the long edge is set to base_resolution, the short edge is proportional, and both are rounded down to a multiple of 8. That rounding is why you get clean numbers like 1024×576 instead of 1024×438.86 - they drop straight into EmptyLatentImage.

The real magic is in the web frontend. The JSON is hidden behind a drag-sortable card grid with an aspect dropdown per card, so you reorder shots by dragging instead of hand-editing a blob of text. The node accepts plain pipe-separated text too (name|prompt|aspect), so even if the JSON widget gets mangled in a workflow file, it still parses.

The inputs that matter

• frames_data - the shot list itself. Multi-line JSON, one object per frame. This is what you'll spend your time on.
• select - which frame is active. Matches by name (case-insensitive) or by index; default is "0". The UI renders this as a dropdown of frame names.
• base_resolution - the long edge in pixels, default 1024, range 512–2048. This sets the "size" of your whole storyboard.
• default_aspect - the fallback aspect for frames that don't set their own. 13 presets from 21:9 to 9:21.
• columns - display-only. It controls how many cards per row in the node's grid and doesn't touch the math.

The outputs that matter

FrameGrid is a single-frame node, so pick one frame and read its data:

• prompt, name, aspect_ratio - the active frame.
• width, height - INT, multiples of 8, ready for EmptyLatentImage.
• frame_index, frame_count - where you are in the sequence.
• all_prompts - every prompt joined with \n---\n, handy if you want a text overview of the whole board in one string.

Wire width/height → EmptyLatentImage, prompt → CLIP Text Encode, and you've got one shot rendered. The pack's example workflow does exactly this with SD1.5 and expects v1-5-pruned-emaonly.safetensors.

Install

Two ways, both painless. The pack has no dependencies - requirements.txt is effectively empty and nothing installs into your Python env, which is a rare treat given the usual dependency hell in this ecosystem.

ComfyUI Manager - search "comfyui-storyboard-suite" and install, then restart:

cd ComfyUI/custom_nodes
git clone https://github.com/Smyshnikof/comfyui-storyboard-suite

Restart ComfyUI, then right-click → Add Node → Storyboard. No models to download.

Gotchas

• select binds by name. Drag-sort reorders the cards, but the active frame follows its name, so reordering won't silently swap the frame you're looking at. The flip side: rename a frame and a selection pointed at it quietly falls back to the first frame.
• Because dims round down, a 21:9 frame at base 1024 comes out 1024×432, not the exact 438.9 - close enough, but don't expect pixel-perfect ratios.
• It's a brand-new, small pack. If the fancy panel doesn't render after install, hard-refresh the browser; the raw JSON widget underneath still works, so you're never fully stuck.