The Asset Style Guide Generator
Last reviewed
IntermediateWhat you'll learn
~30 min- Plan and build a Node.js CLI tool that produces a complete visual style guide as HTML
- Define art direction for characters, environments, UI, and VFX in written form
- Create ready-to-use AI image generation prompts for Midjourney, ChatGPT Images and Stable Diffusion
- Output a print-ready HTML style guide document suitable for sharing with artists and collaborators
What you’re building
You have a playable combat prototype with movement, enemy AI, and juice. The mechanics prove the concept. But when you show it to someone — a potential collaborator, an artist, an investor — the colored rectangles on a black background do not sell the vision. They ask: “what is it supposed to look like?” And right now your answer is “well, imagine BloodRayne but…” followed by waving your hands.
A style guide answers that question concretely. It defines the visual identity of your game: what Rayne looks like (red hair, updated leather outfit, dual blades), what the world looks like (gothic architecture, dark moody lighting, crimson and charcoal palette), what the UI looks like (minimal HUD, blood-red accents), and what the effects look like (visceral particles, dramatic lighting). More importantly, a style guide gives AI image generators specific prompts so you can create concept art without hiring an artist yet.
In this lesson you are building a Node.js CLI that generates a complete visual style guide as a styled HTML document. You define your game’s aesthetic vision in structured data, and the tool outputs a print-ready document with written descriptions and copy-paste prompts for Midjourney, ChatGPT Images and Stable Diffusion.
This is the same pattern as the RFP machine or the resident response studio — structured input transformed into a formatted professional document. In this case, the input is art direction data (character descriptions, color palettes, mood references) and the output is an HTML style guide. This pattern works for any brand guide, design system documentation, or creative brief.
The showcase
Here is what the finished CLI produces:
- HTML style guide with five sections, styled with CSS for professional presentation
- Character art direction — written description of Rayne (modernized design, signature red hair, combat outfit), with generation prompts for each image tool
- Environment palette — color swatches (gothic dark palette), mood descriptions, reference notes, with generation prompts for key environments (cathedral, underground lab, rain-soaked rooftop)
- UI style — HUD design language, font direction, menu aesthetic, with generation prompts for UI mockup screens
- VFX style — particle aesthetics, color theory for effects (blood = deep crimson, energy = purple-red, dodge = spectral blue), with generation prompts for effect reference sheets
- Color palette page — hex codes displayed as visual swatches with labels and usage notes
- AI prompt appendix — all generation prompts collected in one section, organized by tool (Midjourney, ChatGPT Images, Stable Diffusion), ready to copy and paste
The HTML output is self-contained — open it in a browser and it renders as a clean, professional document. Print it to PDF and it looks like something a studio art director produced.
Before you start
If your AI tool is open from the last lesson, leave it (/exit, or Ctrl+D in Antigravity) so you are back at your terminal prompt. Create a folder for this generator and step into it:
mkdir -p ~/retro-remake/14-asset-style-guide-generator && cd ~/retro-remake/14-asset-style-guide-generatorIn Windows PowerShell, run mkdir ~/retro-remake/14-asset-style-guide-generator; cd ~/retro-remake/14-asset-style-guide-generator. In a desktop agent app, open ~/retro-remake/14-asset-style-guide-generator as your workspace folder. Then start your AI tool the way Workbench Setup showed.
Switch to plan mode
Switch into Plan mode (Shift+Tab, /plan, or the mode picker) so your agent asks about the visual sections, palettes, and image-generator formats before writing code.
The conversation
Each message below is an example. Say it in your own words.
Say what you want
Ask for a generator that builds a self-contained HTML visual style guide and prompt appendix for your BloodRayne remake:
Build a tool that generates a self-contained HTML visual style guide for a BloodRayne remake called RAYNE, with color swatches and AI image prompts. Ask me what you need first.Answer its questions
Your agent will ask about the sections in the guide, the environments and palettes, which AI image tools to format prompts for, and the output options:
| If it asks | You might say |
|---|---|
| Which sections should the style guide include? | Five sections: Rayne character design, three environments, UI style, VFX style, and a complete prompt appendix. |
| What are the three environments and their mood? | A ruined Gothic Cathedral, a cold Underground Laboratory, and a neon Rain-Soaked Rooftop, each with five hex swatches. |
| How should Rayne and the UI be described? | Deep red hair, tactical crimson-and-black leather, forearm blades, plus a minimal diegetic HUD with blood-red accents. |
| Which AI image generators should the prompt builder target? | Format 15 to 18 prompts three ways: Midjourney (--ar 16:9), ChatGPT Images prose, and Stable Diffusion tags with negatives. |
| What options and output formats should the generator support? | Let me customize the title, subtitle, and output folder, export HTML or JSON, and print cleanly to PDF. |
Read the plan
Before approving the build, check the agent’s plan for these four details:
- Rich structured art direction (
src/art-direction.js): includes Rayne’s physical description, tactical leather outfit, wrist-mounted blades, action poses, and 5-swatch palette (#8B0000,#1a1a1a,#2d1f1f,#c0c0c0,#ff4444), plus descriptions, lighting notes, and 5 hex swatches each for the Gothic Cathedral, Underground Laboratory, and Rain-Soaked Rooftop, as well as UI and VFX palettes. - Three distinct prompt formats (
src/prompt-builder.js): generates prompts for character turnaround sheets, action poses, wide and detail environment shots, HUD mockup, and VFX reference sheet in Midjourney (--ar 16:9, no hardcoded version flag), ChatGPT Images (natural sentences with mood and hex palette), and Stable Diffusion (comma-separated tags plus| Negative:exclusions). - Self-contained HTML with real CSS color swatches (
src/html-builder.js): renders a dark-themed cover page, table of contents, inline colored swatch boxes (background-colorset to each hex code), copyable prompt blocks, and@media printrules for clean PDF export. - CLI options (
src/index.js): accepts--title(default"RAYNE"),--subtitle,--output(default./output),--format(htmlorjson), and--dark, using CommonJS withchalk@4.
If the plan outputs plain hex text instead of rendered color swatch blocks or uses identical prompt strings for all three image tools, correct it before approving:
Show every palette color as a filled visual swatch box next to its hex code, and format Midjourney, ChatGPT Images, and Stable Diffusion prompts differently.Once the plan covers all four points, approve it and let your agent build the generator. After opening output/style-guide.html in your browser, ask your agent to save a Git checkpoint.
Visual identity requires concrete detail. “Gothic and dark” produces generic results. “Ruined gothic cathedral with broken stained glass casting fractured colored light across stone floors, blood-red banners hanging from pillars” gives both human artists and image generators a clear target.
What you get
After the AI finishes, the project folder structure looks like this:
14-asset-style-guide-generator/ package.json src/ index.js art-direction.js prompt-builder.js html-builder.js output/Generate and view
Ask your agent to install what is needed, generate the style guide, and open it:
Install any needed packages, run the style guide generator, and open output/style-guide.html in my browser.Or run it from a terminal:
cd ~/retro-remake/14-asset-style-guide-generatornpm installnode src/index.jsopen output/style-guide.html # macOS (Linux: xdg-open output/style-guide.html; WSL: explorer.exe output then double-click style-guide.html)The HTML file opens in your browser as a dark-themed professional document. Scroll through it. The cover page has your game title. The character section describes Rayne with color swatches rendered as actual colored blocks. The environment section paints three distinct locations. The prompt appendix has every AI image prompt numbered and ready to copy.
Print it to PDF (Cmd+P or Ctrl+P, Save as PDF). That PDF is something you can hand to an artist, attach to a pitch email, or pin to your wall as a reference.
If something is off
| Problem | Follow-up prompt |
|---|---|
| Color swatches show as text, not colored blocks | The color palettes only show hex codes as plain text. Display a visible colored swatch block right next to each hex code and usage label so an artist can see the palette at a glance. |
| HTML file is unstyled | The generated style guide opens as plain unformatted text. Style the HTML document with a dark gothic background, crimson section headings, clean readable typography, and dark callout boxes for the copyable image prompts. |
| Prompts are identical for all three AI tools | The Midjourney, ChatGPT Images, and Stable Diffusion prompts all use the exact same wording. Format each tool differently: widescreen aspect flag for Midjourney, natural descriptive prose for ChatGPT Images, and comma-separated tags with negative prompts for Stable Diffusion. |
Deep dive
Why a style guide before any art exists
Art is the most expensive part of game development. An artist charges per asset, and revisions cost more than initial creation. A style guide reduces revisions by defining the visual target before anyone draws a single pixel. When you tell an artist “dark gothic with red accents” they interpret that through their own taste. When you hand them a document with specific hex codes, lighting descriptions, material notes, and reference images (generated from your AI prompts), they execute your vision on the first pass. The style guide pays for itself in the first round of avoided revisions.
AI image prompts are a workflow, not a replacement
The prompts in this style guide are not meant to produce final game art. They are meant to produce reference images — mood boards, color studies, composition ideas — that communicate your vision to human collaborators (and to yourself). When you paste the Rayne character prompt into Midjourney and get back an image that is 70% right, that 70% is more useful than a verbal description. You can annotate it: “like this but the hair is longer” or “the blade angle should be more curved.” Reference images accelerate communication. That is their job.
The three AI tools want different things
Midjourney tends to reward evocative, artistic language with style modifiers (cinematic lighting, concept art, high detail). ChatGPT Images tends to reward clear, descriptive natural language with explicit mood and color instructions. Stable Diffusion front ends tend to reward comma-separated tags and a negative prompt. The prompt builder generates all three formats from the same source data because you do not know which tool will produce the best result for a given concept. Try all three. Use what works.
🔍The style guide as a living documentclick to expand
This style guide is generated from a JavaScript data file. That means it is versioned, diffable, and updatable. As your vision for the BloodRayne remake evolves, you update art-direction.js and regenerate. The old style guide is replaced with the new one. If you are collaborating with others, the art direction data can live in your git repository — every team member sees the same visual target, and changes are tracked in commit history. This is better than a Figma file that someone forgets to update.
Customize it
Add a mood board grid
Add a --moodboard flag that generates output/moodboard.html with a responsive 12-to-15 cell grid containing each section label, a copyable Midjourney prompt, and a 400x250 placeholder box tinted with that section's primary color.Add a competitor reference section
Add a Visual References section comparing BloodRayne 1 and 2, Devil May Cry 5, Symphony of the Night, Hollow Knight, and Dead Cells, with one sentence each on what to borrow and avoid.Add a sprite sheet specification page
Add a Sprite Specifications section defining 64x96px character frames, frame counts for idle, run, combo, dodge, hurt, and death, PNG alpha export rules, file naming conventions, and a visual grid diagram.Try it yourself
- Walk through the plan-mode conversation above and generate the style guide.
- Open
output/style-guide.htmlin your browser and read through every section. Does Rayne’s description match the character in your head? If not, editart-direction.jsand regenerate. - Copy one of the Midjourney character prompts. If you have Midjourney access, paste it and see what comes back. If not, paste the ChatGPT Images prompt into ChatGPT and ask it to make the image. Compare the result to your mental image.
- Print the HTML to PDF. Would you feel confident handing this document to a freelance character artist? If not, what is missing? Add it to the data file.
- Try the
--format jsonflag and look at the raw data output. This JSON can be imported by other tools — your future level editor or asset pipeline could read it. - Change the game title with
--title "BLOOD HEIR"or whatever your working title becomes. Regenerate. The entire document updates.
A game’s visual identity is a decision you make once and live with for years. This style guide is how you make that decision concretely, communicate it clearly, and evolve it systematically. Rayne has waited long enough to look good again.
You now have a professional style guide document that defines what your BloodRayne remake looks like — from Rayne’s updated design to the gothic environments to the particle effects. You have AI image prompts ready to generate concept art. You have a printable document to share with collaborators. The colored rectangles in your Phaser prototype are temporary — the style guide defines what replaces them. Next lesson: building the HUD and inventory systems that tie the gameplay together.