TS Style Prompt Selector

Everyone has had the moment: you know you want that look, but the words won't come. "It's like an old print… with the texture… you know?" The TS Style Prompt Selector is the pack's answer - a visual style picker with a library of 157 styles, each with a thumbnail preview, grouped into 15 categories that run from cave painting and Byzantine mosaic through the twentieth-century avant-garde and film noir to pixel art and vaporwave. You browse pictures until you find the look, click, and get the matching prompt text out.

It's the difference between knowing a style by name and being able to find a style you can't name yet. That second one is the whole reason this node exists.

How the output is designed

The trick is in what comes out. Each style is a pure style modifier - medium, technique, palette and texture, with no subject of its own - and it deliberately ends with a comma and a space. That trailing comma is not an accident. The output is meant to be prepended to your own prompt, and the comma makes the two halves read as one sentence:

style   ukiyo-e woodblock print style, bold black keyblock outlines, visible paper grain,
yours   portrait of an old fisherman
result  ukiyo-e woodblock print style, bold black keyblock outlines, visible paper grain, portrait of an old fisherman

In the graph, that means a concatenate node with this node's prompt in string_a and your prompt in string_b. If no style is chosen, the output is a blank space - so the chain keeps working even before you pick one.

The mechanism is simple under the hood: the style_id input selects an entry, and the prompt text itself lives in styles.json, not in the node's inputs. The pack even fingerprints that file so editing a style's text invalidates the cache - you don't get stale prompts after tweaking a definition.

The interface

• style_id - the id or name of the style to load. In the studio this is a browsable picker; by hand you can type the id directly.
• prompt (output) - the style modifier as a STRING.

Names, categories and descriptions are bilingual and follow the ComfyUI interface language (English or Russian).

Installing it

It ships in comfyui-timesaver: ComfyUI Manager → search Timesaver, or

cd ComfyUI/custom_nodes
git clone https://github.com/AlexYez/comfyui-timesaver
cd comfyui-timesaver
python -m pip install -r requirements.txt

then restart ComfyUI.

Gotchas

The output is a modifier, not a full prompt - feed it to a text encoder on its own and you'll get a subjectless generation. Chain it before your real prompt, comma included. Also worth remembering: on the LLM-encoded model families that dominate in 2026, "in the style of X" phrasing is exactly the kind of language an instruction-following encoder actually responds to, but these modifiers were written with SDXL-era prompt structure in mind, so sanity-check the first result rather than assuming one click nails the look. It's a fast way to find a style; the final wording is still your call.