◊ Scenes
Scene and background auto-writer
- PROMPT
Describing a scene well is genuinely tedious in tag form - time of day, weather, ambience, props, whether there's a crowd, what the subject is doing - and most people either skip half of it or write the same three scene tags every time. IllustriousScenesPlus is a scene generator: pick a category and a complexity level, flip on the categories of detail you want, and it assembles a scene description in Danbooru-tag style so you don't have to.
How it decides what to write
Category (11 choices - Daily Life, Outdoor, Indoor, Seasonal, Atmospheric, Action, Emotional, School Life, Fantasy Adventure, Romance, Convention) sets the vocabulary pool. Complexity (simple/medium/detailed) controls how many scene elements actually get pulled in. From there, a set of Include * booleans decide which categories of detail show up: Time/Weather, Ambience, Event, Prop, Density (crowd level), Person Description, Pose/Action, and Clothing.
That last one is worth calling out: if Include Clothing is on, this node exposes the exact same Outfits/Top/Bottoms/General Style/Headwear/Hair Colors dropdowns the standalone IllustriousClothing node has, so you can describe an outfit here without wiring in a separate node. Safe Adult Subject swaps girl/boy for woman/man phrasing. Danbooru Tag Style forces lowercase, underscore-separated output to match the convention Illustrious was actually trained on. Token Count (77/150/250/300/500) is a soft budget the node prunes toward, since Illustrious runs on SDXL's 77-token CLIP pair by default and anything past that starts getting truncated or ignored.
Enable TIPO Optimization turns on candidate generation and ranking - the node builds several prompt variants (TIPO Candidates, 3–32) and picks the best one for a given TIPO Flavor (balanced/vibrant/soft/natural), respecting a TIPO Max Length character cap. This is the same optimization logic that lives in the standalone TIPOPromptOptimizer node - if you already understand that node, you understand this toggle.
Optional prefix/suffix strings and tipo_extra_negatives (terms to downweight during optimization) round out the inputs. Output is a single PROMPT string.
Installing it
ComfyUI Manager: search "EasyIllustrious," install, restart. Manual: cd ComfyUI/custom_nodes && git clone https://github.com/regiellis/ComfyUI-EasyIllustrious, then pip install -r ComfyUI-EasyIllustrious/requirements.txt inside your ComfyUI environment, then restart.
The gotcha worth knowing about before you chain this in
A user on the pack's launch thread got confused finding unexpected tags like beauty_mark and fireworks showing up in their final prompt with no idea where they came from - it turned out a scene generator node was wired into their prefix/suffix chain and quietly injecting its own picks every run. The author's own fix was blunt: disconnect the scene generator from the suffix input if you want full manual control over your prompt. It's worth remembering with this node specifically, since its whole point is auto-injecting scene content - if you're building a prompt by hand and just want this node for one specific piece (say, only the weather tag), turn off every Include * toggle you don't want rather than leaving them on defaults and hoping the output matches what you expected.
Inputs (30)
| Name | Type | Default | Description |
|---|---|---|---|
| Category | COMBO | Outdoor | 11 options: Daily Life, Outdoor, Indoor, Seasonal, Atmospheric, Action, +5 |
| Complexity | COMBO | medium | 3 options: simple, medium, detailed |
| Include Time/Weather | BOOLEAN | true | — |
| Include Ambience | BOOLEAN | true | — |
| Include Event | BOOLEAN | false | — |
| Include Prop | BOOLEAN | true | — |
| Include Density | BOOLEAN | false | — |
| Include Person Description | BOOLEAN | false | — |
| Include Pose/Action | BOOLEAN | true | — |
| Include Clothing | BOOLEAN | false | — |
| Outfits | COMBO | - | Pick a full outfit or leave '-' for random. |
| Top | COMBO | - | Pick a top; used if no Outfit is chosen. |
| Bottoms | COMBO | - | Pick bottoms; used if no Outfit is chosen. |
| General Style | COMBO | - | Overall clothing style/aesthetic. |
| Headwear | COMBO | - | Specific headwear item to include. |
| Hair Colors | COMBO | - | Force a specific hair color or leave '-' for random. |
| Safe Adult Subject | BOOLEAN | true | Use woman/man instead of girl/boy tags. |
| Use Chain Insert | BOOLEAN | true | — |
| Strict Tags (no phrases) | BOOLEAN | true | — |
| De-duplicate With Prefix/Suffix | BOOLEAN | true | — |
| Danbooru Tag Style | BOOLEAN | true | Format output to lowercase_underscore Danbooru tags. |
| Token Count | COMBO | 77 | Approx. token budget for the prompt |
| Enable TIPO Optimization | BOOLEAN | true | Generate prompt candidates and pick the best for Illustrious. |
| TIPO Candidates | INT | 83–32 | Number of variants to consider. |
| TIPO Flavor | COMBO | balanced | Illustrious-oriented flavor to bias scoring. |
| TIPO Max Length | INT | 32080–800 | Soft cap for final prompt length (characters). |
| TIPO Seed | INT | 00–2147483647 | Randomization seed (0 = auto). |
| prefixopt | STRING | Existing text before scene (chain-aware). | |
| suffixopt | STRING | Existing text after scene. | |
| tipo_extra_negativesopt | STRING | Comma-separated terms to downweight/remove during TIPO optimization. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| PROMPT | STRING | — |