Apex Prompt

Apex Prompt Preset is the node for people who keep typing "golden hour" and getting a flat beige nothing. It hands you four curated dropdowns - Environment, Lighting, Style, and Camera Lens - and stitches whatever you pick onto your own prompt text, ready to feed a text encoder. Think of it as a library of well-written scene descriptions you can borrow instead of writing from scratch. For a beginner this is genuinely useful: describing an environment and a lighting setup in a way diffusion models actually respect is a skill, and the presets skip the learning curve.

The mechanism is refreshingly boring. The node ships with a prompt_presets.json file holding all the preset phrases - things like "Modern subway train interior, stainless steel poles..." for environment and "Volumetric Moonlight" for lighting. On load it reads that file, and when you run it, each dropdown selection (or "Random") contributes its text to a combined output. The seed input makes "Random" reproducible: same seed, same selections, every time. It's deterministic prompt engineering, not an AI writing your prompt for you - which is the point. You stay in control, the presets just save you the effort of phrasing.

The inputs that matter

• input_text - your own prompt, which gets prepended (or combined) with whatever the presets produce. Can be empty if you're going full preset.
• seed - locks in the "Random" choices so you can regenerate the same combo.

The four dropdowns are each "Disabled" (off), "Random", or a specific preset - Environment has 34 choices, Lighting 31, Style 14, and Camera Lens 45. The lens list is the fun one: real focal lengths like 35mm, 50mm, 85mm, plus cinema and specialty lenses, each dropping a phrase like a "35mm street" description into your prompt. Outputs are the full combined_prompt plus separate environment_text, lighting_text, style_text, and camera_lens_text strings, so you can wire just the lighting phrase somewhere else if you want.

The honest take

This is a crutch with taste. The presets are genuinely well-crafted phrases - the "Cinematic Elegance" and "Photorealistic" style entries read like something a working artist would actually write. But preset prompts have a ceiling: everyone using this node uses the same phrases, so your outputs will drift toward a common look, and nothing here teaches you why "volumetric moonlight" works. Use it to get results fast, then read what it produced and steal the structure for your own prompts. Also note the presets are plain strings - they don't adjust sampler settings, CFG, or anything else, so a "Cinematic" preset changes your text, not your pipeline.

Installing it

Part of the Apex Artist pack. ComfyUI Manager → search "Apex Artist" → Install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/ApexArtist/comfyui-apex-artist.git

Restart ComfyUI; it's under Add Node → Apex Artist → Text. No model downloads, no extra dependencies. The prompt_presets.json file lives inside the pack folder - you can edit it to add your own presets, though you'll want to back it up, because a future pack update can overwrite it.

Gotchas

If the dropdowns ever come back empty, the JSON likely got corrupted or overwritten - the node falls back to built-in defaults, so it recovers gracefully. And because "Random" is seed-driven, your output changes completely between seeds, which trips people up when they expect a stable prompt from a preset workflow. Wire combined_prompt into a CLIP Text Encode node and you're done - that's the whole integration.