NA - Get Profiles

NA - Get Profiles is the most honest node in the Neural Atelier pack: it doesn't call any API, it doesn't need a Gemini key, and it can't spend a cent of your Google bill. It's a glorified directory listing. Given a prompt pack, it reads the pack's configs/ folder and hands you back the names of every prompt profile inside it, comma-separated, on a single STRING output.

So what's it for? Discovery, mostly. The flagship NA - Sketch to Photo Orchestrator ships 42 built-in profiles - sketch-to-photo, ghost mannequin, hanger, 2x2 grid, editorial beach attitude, and so on - and unless you memorize the config folder, you have no way to know what's installed without opening Explorer. This node is that answer, inside the graph. It's also genuinely useful if you're building a dynamic workflow, a script, or a frontend that needs the profile list at runtime instead of hardcoding names.

How it works: the pack ships three prompt packs as folders under configs/ - NA_Recolor, NA_Sketch_to_Photo_Orchestrator, and NA_Styling_Detail_Change - each full of .json master-prompt files. GetProfiles lists the JSON filenames in whichever pack you pick and joins them with commas. It's the exact same loader the SketchToPhoto node uses to populate its prompt_profile dropdown, so the two can never disagree. No model downloads, no dependencies beyond the pack itself, no network.

The inputs and output that matter:

• prompt_pack (required, dropdown) - one of the three packs. That's the entire input surface.
• profiles (output, STRING) - the comma-joined profile list. Wire it into a text display node (Show Text / anything that takes a STRING) and read it off.

That's the whole node. It's a tool for you and for workflow authors, not a renderer, so don't expect output images from it.

Install: same as the rest of the pack. In ComfyUI Manager, search "Neural Atelier" and install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/valsecchi75/comfyui_neural_atelier
cd comfyui_neural_atelier
pip install -r requirements.txt

Restart ComfyUI. Dependencies are light - google-genai, Pillow, torch, numpy - and there are no weights to download because everything runs through Google's cloud.

Where people get tripped up: the output is comma-joined, so the 42-profile sketch-to-photo pack returns one very long run-on string - that's ugly but expected, not a bug. And remember, the node only lists profiles for packs that actually exist in your install; if you deleted a config folder, it just won't show up. It's the closest thing Neural Atelier has to a "what can this pack do" cheat sheet, and for a beginner staring at an empty canvas, that's a surprisingly handy place to start.