Select latent size 2Mp

Full credit to the author for honesty: the pack's own README introduces this node as added "for testing purposes" and admits they "don't think anyone needs it" unless a use turns up. That's a refreshingly straight framing, and it's also slightly out of date - a 2-megapixel dropdown genuinely does line up with where a chunk of current-generation models actually run. Several 2026-era models (Z-Image, Flux 2 Klein among them) natively operate in a 1–2MP band rather than SDXL's fixed 1MP, so a named list of resolutions up around 2MP isn't the dead end the README suggests.

It's the same idea as its sibling Select Latent Size 1Mp, just with every entry roughly doubled in area: "Square (1448x1448) 1x1," "HD Video landscape (1920x1080) 16x9," "Cinema landscape (2176x960) 21x9," and their portrait mirrors - seventeen named entries in total.

How it works

Pick a named resolution from the dropdown - landscape and portrait are listed separately, so there's no swap toggle to hunt for, just the mirrored entry further down the list. batch_size stacks copies in the output batch.

Inputs and outputs that matter

• resolution (default Square (1448x1448) 1x1) - the named size to build the latent at.
• batch_size (1–10, default 1) - how many copies to generate.

One output: Latent, sized and batched, ready for a sampler.

Installing it

ComfyUI Manager: search mzmaxam in Custom Nodes Manager and install. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/MzMaXaM/ComfyUi-MzMaXaM

Restart ComfyUI. No models or extra dependencies to worry about - clone and restart covers it.

Where people get tripped up

This is the least battle-tested node in the pack by the author's own admission, so treat the list as a reasonable starting point rather than a definitive one - if a size you need isn't on it, there's no way to type a custom value here. For that, use the pack's Select Latent Size Plus node instead, which covers a much wider resolution range and also gives you numeric width/height outputs this node doesn't have. More importantly: check that your checkpoint actually wants ~2 megapixels before reaching for this list. It's a good fit for the newer 1–2MP-native models, but a poor one for SDXL-family checkpoints (Pony, Illustrious, NoobAI), which were trained around 1 megapixel - going noticeably above a model's trained resolution is a common source of duplicated elements and soft, degraded detail rather than a genuinely sharper image. If you're on an SDXL-family model, Select Latent Size 1Mp is almost certainly the node you actually want. And the usual batch note applies: batch_size scales VRAM and time linearly, not quality.