Stack_text

Stack_text is the prompt-entry node in Apt_Preset's ecosystem - but it's carrying a lot more than just two text boxes. Alongside pos (multiline) and neg for your actual prompt text, it ships a style dropdown with over 300 preset entries, the same idea as the big SDXL-style-JSON packs you've probably seen bundled with other UIs (Fooocus does this too), where picking a style silently appends genre- or medium-specific prompt fragments on top of whatever you typed.

How the style presets are organized. They come in named clusters - general-purpose basics, realism, "effect image" looks, a pure/clean category, then a large block sourced from Stability AI's own official style list (3D model, simulated film, animation, cinematic, comic, clay, digital, and more), plus over 300 total entries the schema only samples a handful of. If you've used any Fooocus-derived style picker before, the mental model transfers directly: pick a style, it quietly does prompt engineering for you underneath your own words.

One thing worth knowing before you open that dropdown: most of the category labels are in Chinese rather than English - "通用-写实" for a general realism style, "SAI官方-漫画" for the SAI-official comic style, and so on. The underlying prompt fragments each style injects are presumably fine regardless of language, but if you're expecting an English-labeled menu, it isn't one, and you'll want to either recognize the categories by trial and error or keep a translation open the first few times you use it.

Chaining. Like Stack_LoRA and Stack_Redux, this node has a text_stack optional input for feeding in another Stack_text node's output - letting you layer, say, a base-prompt node into a style-only node into a final combined output, rather than cramming everything into one node's fields.

Output: text_stack, the pack's TEXT_STACK type, meant for a downstream Apt_Preset sampler or loader node that reads and combines the positive/negative/style pieces - not a direct substitute for CLIP Text Encode, since it doesn't encode anything itself.

Installing it. No special dependencies for this node: cd ComfyUI/custom_nodes && git clone https://github.com/cardenluo/ComfyUI-Apt_Preset, restart ComfyUI, or search "ComfyUI-Apt_Preset" through ComfyUI Manager. Run the bundled install.bat on Windows once after cloning - other nodes in this pack need extra Python packages that Manager's own resolver sometimes doesn't pull in automatically.

Troubleshooting. If picking a style doesn't seem to change your output at all, remember this node only builds the text - check that its text_stack output is actually wired downstream into whatever applies the encoding, the same "config node does nothing on its own" pattern that shows up across this pack. If a style's effect feels too strong or too weak relative to your own prompt, there's no strength slider on this node for it - the style fragment gets appended as plain text, so its influence is whatever CLIP naturally weights it as, which means adjusting your own prompt's length and specificity is the practical lever, not a setting on this node. And if you're chaining multiple Stack_text nodes and getting duplicated or conflicting style text, check you haven't accidentally set a style on more than one node in the chain - since styles append rather than replace, stacking two style-set nodes together is additive, not a last-one-wins override.