Prompt Settings Slider

This is the bigger sibling of the Environment Slider - where that one handles where the image happens, Prompt Settings Slider covers how it looks. Twenty-one dropdowns covering shot style, artists, film directors, architects, photography styles, lighting, effects, magic elements, cameras, films, image type, composition, art styles, themes, printing materials, illustration style and color theme. Pick from the lists, get a finished prompt string. It's the kind of node that looks intimidating on first open and then quietly replaces half your prompt-block discipline.

How it works

The node loads a bundled settings.json and builds one dropdown per category. Every dropdown follows the same pattern: None to skip, a named choice, or Random to roll one entry. Each category has a weight slider (0.1–2.0) - set it to 1 and the tag goes in plain; anything else and it's wrapped as (tag:weight), which is the classic CLIP-era attention syntax.

The lists are the selling point. Artists has 154 names, Art Style and Illustration Style have 99 and 116 entries each, Shotstyle 1/2 share a 106-entry lens-and-angle list (14mm wide angle, dolly zoom, Dutch angle...), and Film Directors covers 72 of them from Kurosawa to Christopher Nolan. It also includes the occasional delight like Arists Special - yes, spelled exactly like that, it's the author's typo baked into the input name - a curated 21-name shortlist (Chris Leib, Frank McCarthy, Jamie Hewlett...).

The inputs that matter

The full list: Shotstyle 1/2, Artists, Arists Special, Film Directors, Architects, Photography Styles, Lighting, Effects, Magic Elements, Camera, Films, Image Type, Composition Method, Art Style, Art Style+, Scene Atmosphere, Theme, Printing Materials, Illustration Style, Color Theme - each with its weight slider. The seed is there but, same as the Environment Slider, the Random picks don't actually consult it in the current source; the roll comes from Python's global random. Specific values if you need reproducibility.

Everything selected gets comma-joined and the node appends a trailing comma, so the output prompt string wires straight into CLIP Text Encode.

Which model is this for?

Honest take: this is an SDXL-era tag prompt builder. The comma-joined booru-style tags are exactly what Illustrious, NoobAI and Pony want. On Flux the tags still mostly translate, but (tag:1.3) weighting is discarded by Flux's T5/LLM-style encoder, so the sliders silently stop mattering. Don't fight it - if you're on a modern LLM-encoded model, this node's value drops to just "structured block of descriptive words," which is still more than nothing.

Installing & the catch

Via Manager, search ComfyUI-RvTools_v2; or git clone https://github.com/r-vage/ComfyUI-RvTools_v2 into ComfyUI/custom_nodes and restart. It shows up under 🫦 RvTools II → Text, no model downloads. The pack's README says v2 is no longer maintained (superseded by ComfyUI_Eclipse) but the node is simple and stable. And remember: v2 renamed every node from the original v1, so old workflows referencing ComfyUI-RvTools nodes will show missing - that's expected, not a bug in your install.