凤希AI - 宽高设置

FxAiPhotoSizeConfigV2 is what the pack's size-config node looks like after the author listened to feedback. The original FxAiPhotoSizeConfig had a 尺寸设置 boolean that switched between "use my typed width/height" and "parse the preset". In practice that toggle confused people - flip it the wrong way and the preset you picked gets ignored, or worse, your custom values get overridden. V2 deletes the toggle entirely. Now the preset just writes the width and height straight into the boxes, and what you see is what gets output. One less failure mode, same job.

It lives in 凤希AI/工具 (Tools). The inputs are 预设 (a dropdown of 11 presets: 1寸 through 10寸 photo prints plus 正方形 1024×1024, 竖屏 1080×1920, and 竖屏 1440×1920), 宽度 and 高度 as INTs (64–8192, step 8), and a 反转 checkbox that swaps the two dimensions before they leave the node. Note the preset list here is slightly different from V1 - this one swaps V1's landscape 1920×1080 for a second portrait option (1440×1920), which tells you the author's audience is mostly portrait video and ID-photo work.

The mechanism is a pure pass-through: whatever 宽度 and 高度 hold at execution time is what comes out, with 反转 applied if you ticked it. That's the whole node. The two outputs, 宽度 and 高度 (both INT), are meant to feed an EmptyLatentImage, a sampler's width/height, or the encode nodes in this pack's Qwen workflows - anything that wants an explicit resolution in multiples of 8.

A small thing that matters: because the front-end writes the preset value into the widgets, the "source of truth" is the width/height boxes, not the dropdown. If you're building a workflow where another node overwrites the width port, the preset won't fight it - which is exactly the behavior the toggle in V1 made awkward. For scripted or API-driven graphs this is the better-behaved of the two versions.

If you only need one of these, pick V2 - it's the same preset set plus the 1440×1920 portrait size, minus the confusion. It ships inside fxai-toolkit (凤希AI, fxai.site), installed the same way as every node in the pack: ComfyUI Manager search "fxai-toolkit", or

cd ComfyUI/custom_nodes
git clone https://github.com/fxai666/fxai-toolkit

then restart ComfyUI. The pack pulls in soundfile and psutil automatically on first load, and all the 凤希AI nodes group under the 凤希AI menu. It's a small utility, but if you're running the pack's ID-photo or portrait-video flows, it's the dimension source they're pre-wired to expect.