AC_Fifth_Season提示词

The fifth season of ComfyUI-AC_FUNV2.0 is where the pack's prompt library starts leaning hard into 写实 - photorealism. AC_Fifth_Season提示词 ships 100 prompts for realistic women: 写实广告系列 (realistic advertising series), 写实温馨小屋 (a cozy cottage interior), 写实职场OL (office lady), 写实蝴蝶梦幻少女 (dreamy butterfly girl), and a string of beauty/portrait styles. This is the season you reach for when you want an image that's supposed to read as a photo, not an illustration.

Before the mechanism, the content flag you should have: read the filenames before you use them, because this is the pack's NSFW-adjacent shelf. Several files are titled around 性感 (sexy) and 纯欲风 (a "pure desire" aesthetic that's exactly what it sounds like) - 写实性感包臀裙, 写实性感丝袜猫女, 写实纯欲风. The prompts are nothing more than text files and none of it is enforced, but if you're sharing a screen or generating in a shared space, skim the list first. The pack's own marketing (a Chinese Bilibili teacher's toolbox) just calls them 写真-style prompts, so don't expect a content warning from the source.

Mechanically, same as every season: ModeList lists the .txt files in the Fifth_season folder, you pick one, and the node reads it and emits the contents on a single STRING output. Wire into CLIP Text Encode and generate.

The syntax reality check, which matters more here than in the anime seasons: realistic subject matter exposes the limits of SDXL-era tag soup fastest. Open a file and you'll find the pack's usual ((masterpiece)), best quality, 1girl scaffolding - vocabulary that does nothing on Flux and even on SDXL is mostly wasted on photorealism, where the model and the camera terms do the work. The genuinely useful bits in these files are the photography cues - lighting, lens, framing - which is where the author actually put in effort. If you're on a photoreal SDXL finetune (Juggernaut, CyberRealistic lineage), these will behave. On a modern model, mine the files for the camera descriptors and rewrite the rest as sentences.

The input that matters: ModeList - one required dropdown, no other settings.

Install is the pack standard:

cd ComfyUI/custom_nodes
git clone https://github.com/A719689614/ComfyUI-AC_FUNV2.0

restart, then under ♑AC_FUNV2.0. No models to download, no dependencies beyond stock ComfyUI (the torch==2.11.0 pin in requirements.txt is noise; skip it if pip complains).

Final thought: the photorealism KB keeps saying the same thing this pack's existence demonstrates from the other side - for realistic output, the prompt is a smaller part of the equation than the checkpoint and the sampler settings. These files will get you a starting composition and a camera setup; the model you pair them with decides whether the result looks like a photo or a render. Wire the output into the Panel node first so you can see what you're actually feeding the sampler, pick your checkpoint to match the content, and treat the 100 files as a reference shelf rather than a recipe.