ER Wedge Sequencer Video

ER Wedge Sequencer Video is the frame-batch twin of ER Wedge Sequencer, and it exists because comparing video settings by hand is even more miserable than comparing images. You can't just glance at two thumbnails - you need to watch the motion. If you've ever burned an afternoon re-running an LTX or Wan clip to decide whether 8 steps or 12 steps flickers less, this is the node that ends that.

Same idea as the still version, one sentence: you point it at a video-producing workflow (Create Video, LTX Video, Wan, whatever), wedge a widget on any node in that graph - steps, cfg, the denoise, even a motion seed - give it a value list or range, and it queues a run per value, collects the results into a grid, and plays each one back labelled with the values that made it.

What's different from the image sequencer

Graph-wise it swaps images for a video input and a matching video output, and it passes the video through untouched, so you can keep chaining downstream nodes after it. The optional er_meta string tags runs the same way. Everything else about the wedge definition - ＋ Wedge, ＋ param, the value syntax:

7, 7.5, 8          # list
1:10:2             # range (start:end:step)
rand3              # random values
euler, dpmpp_2m    # combo text values

• is identical to the image node. Multiple params multiply out into every combination, capped at 64 runs.

The real difference is under the hood: each result is a batch of frames, so the node encodes a lightweight MP4 proxy (capped at 448px) for the grid so playback stays smooth, plus an HD proxy (up to 1920px) that's used when you export a contact sheet. The proxy uses the frame rate of your video. And where the still node shows you one image per run, here clicking a result opens the full-resolution file.

The two things that bite people

First, this node depends on PyAV (av). Recent ComfyUI builds bundle it, but if your install predates that you'll get an import error the moment ComfyUI loads the pack - fix it with pip install av in your ComfyUI environment and restart.

Second, the same queue-order rule as the image version: results are matched to runs in queue order, so don't queue unrelated prompts while a video wedge batch is running, or labels land on the wrong clips. Video runs also take real GPU time - a 4-value wedge on a 3-second clip is 12 generations, so set your cap expectations before you ask for 1:30:1.

Installing

It ships as part of the ER VFX Nodepack, not standalone. In ComfyUI Manager search "ER VFX Nodepack" and install, or clone it:

cd ComfyUI/custom_nodes
git clone https://github.com/esperandoelrender/ComfyUI-ER-VFX-Nodepack.git

Restart and search "ER" in the node menu (utils category). If you previously installed the standalone er_* packs this pack grew from, delete those folders first or you'll end up with duplicate nodes. The still-image version of this node needs no PyAV; the video one (this one) does.

And when the wedge results are in and you want to hand them to someone who isn't you - a client, a director, a collaborator - drop ER Wedge Preview Video into the graph to get a clean, labelled gallery with one-click contact-sheet export instead of a wall of node internals.