WanV2V Debug: Number Batch

This pack's entire pitch is "no duplicated or skipped frames across your chunks." NumberedBatchDebugger is how you actually verify that claim instead of trusting it. It stamps a big index number in the top-right corner of every frame in an IMAGE batch, so you can preview a few frames and literally see the sequence.

It's about as simple as a node gets. For each frame it converts to PIL, draws "start_index + i" in the corner, and converts back. White text with a 1px black outline (outline on by default) so it stays readable on both bright and dark footage. All CPU, no model, no VRAM - it's a debug utility and costs you nothing to leave in the graph while you iterate.

The inputs are the four you'd expect: start_index (1) for when you want numbering to begin somewhere other than 1 - handy if you're checking chunk 4 in isolation and want it numbered as though it followed chunk 3; margin (12) for distance from the corner; scale (0.06) for text size relative to the image; and outline (true). Output is the same batch with numbers baked in, so it still passes through to whatever you had downstream.

Where it earns its keep in this pack: drop one on the output of each WanV2VIterControlGroup chunk and one on the merged result from WanV2VMergeDeOverlap. Sequential numbers across the whole merge means your overlap math is right. A jump (…77, 78, 91…) means a chunk lost frames somewhere. A repeat (…81, 81, 82…) means your overlap is too big for the merge's overlap_size, or you fed the merge the wrong batches. This is exactly the "detect overlaps/skips" job the README advertises, and it's genuinely the fastest way to debug a stitch.

One honest warning: it's a debugging node, not a watermarker. Those numbers get burned into the frames, so pull it out of the graph before the final encode - I've seen more than one person ship a render with a "37" floating in the corner. If the text looks tiny or blocky on your system, that's the PIL font fallback kicking in (arial.ttf isn't guaranteed to be present); bump scale to 0.1+ and it'll be legible.

Install is the pack standard, no dependencies beyond what ComfyUI already has:

cd ComfyUI/custom_nodes
git clone https://github.com/Kishor900/comfyui-wanv2v-video-stitcher

Restart ComfyUI and it appears under WanV2vControlnetManager/Debug (ComfyUI Manager: search "ComfyUI WANv2v Video Stitcher"). It's a small node for a narrow job, and it does it without fuss. When your stitching pipeline is behaving, you'll forget it exists. When it isn't, it's the difference between an hour of staring at footage and thirty seconds of reading numbers.