# Style Prompt Studio is a prompt library for generating slides in nineteen looks > This repository is not a program. It is a set of English prompt strings, each describing one visual style for a presentation slide, that you paste into an image generation model to get a 16:9 slide back. It ships as an installable skill for Claude Code, it is MIT licensed, and the styles are written with specific hex colours because the text is meant to be followed literally. **AAAAAAAJ/slides** — slides generates prompts for various expression styles - Repository: https://github.com/AAAAAAAJ/slides - Stars: 532 · Forks: 53 - Language: Python - License: MIT - Published: 2026-10-22 · Updated: 2026-10-22 · Language: en - Canonical page: https://hysenlabs.com/projects/aaaaaaaj-slides ## Nineteen prompts, each a paragraph of visual instructions The repository description is short: slides generates prompts for various expression styles. The readme title is Style Prompt Studio, and the tagline is that it generates multi-style PPT slides with AI, with direct prompts for a model called nanobanana2. What is actually in the repository is text. Each of the 19 styles is a single prompt string in a fenced block, and the strings are specific in a way that matters. The first one, Retro Pop Art, reads: ``` Retro pop art style PPT slide, 1970s magazine aesthetic, flat design with thick black outlines, cream beige background, bold title text, subtitle below, key statistics displayed as cards, Salmon pink #FF6B6B, sky blue #4ECDC4, mustard yellow #FFD93D, mint green #6BCB77 accents, Geometric decorations: quarter circles, concentric rings, star bursts, Bold sans-serif typography, Professional presentation design, 16:9 ``` Every style follows that shape: a named aesthetic, a background, a title and subtitle arrangement, a rule about where key statistics go, a palette with hex values, a decoration vocabulary, a typography direction, and 16:9 at the end. That structure is the point. A prompt like this is not a vibe, it is a specification, and the fact that the palette is given as hex codes means the model has a concrete target rather than an adjective to interpret. ## The first eleven styles, and the eight that were added later The catalogue is split, and the readme is explicit about the split. There are 19 available styles with examples. The note says the first 11 examples are the original Seedream 4.0 outputs, and the added 8 examples use Nano Banana 2 / Pro. It also says that for the new 8 styles, the current English and Chinese entries temporarily reuse the same image. That last clause is the honest bit, and it tells you the examples in the readme are not all current renders. The original eleven were generated with a different model, and for the newer eight the example image is standing in for a translation variant rather than showing a distinct English output. The original eleven cover a wide spread of design languages. Minimalist Clean is described as a white background with generous whitespace, centered title, key stats in simple cards and subtle gray and blue accents. Cyberpunk Neon is a dark charcoal background with a neon glow effect, magenta, cyan and yellow, tech grid patterns and holographic data panels. Neo-Brutalism uses a cream background, bold primary colors and thick 4px black outlines with hard shadows. Swiss International specifies a light gray background, a subtitle with diagonal layout, key stats in geometric blocks and a seven colour palette with Helvetica and asymmetric composition. Others are named for their era rather than their principles. Acid Graphics Y2K uses metallic chrome elements and holographic accents, and Y2K Pixel Retro specifies a pixel font, CRT monitor graphics, isometric tech illustrations and the VT323 font style. Dark Editorial is styled after the New York Times Sunday Review with a black background, a white dot grid and an orange accent. ## The extended styles, built for a single poster-like slide The last eight styles are described differently, and the difference matters for how you use them. They are additional presets optimized for single-slide, poster-like image generation. The original eleven are, in effect, a description of a slide design language that you would apply across a deck, and the extended eight are aimed at one image that has to work on its own. The Bento Grid style is the clearest example. The prompt describes a modular rounded-rectangle card layout filling the full 16:9 frame, an asymmetric but tightly aligned grid composition, a bold title in the primary hero block, a subtitle below, three to five key statistics distributed across smaller cards, and a product screenshot or icon module. The palette is given as off-white, graphite, cobalt blue, mint, soft orange, and the prompt ends by calling it a premium Apple-style product display aesthetic. That layout is a description of a dashboard rather than a slide, and the other extended styles are named for the same kind of artefact: Scrapbook DIY Vibe, and the Neo-Brutalist UI style, which is described as a dashboard interface design with pastel panels, thick 3px black outlines and a card-based layout for a contemporary SaaS dashboard aesthetic. So the extended set is a set of product-screen layouts rather than presentation slides, and picking the wrong group gives you a slide that looks like a dashboard. ## How the workflow is meant to run, with no code at all The quick start is described as requiring no code, and it is three steps. Step one is the prompt: choose any style from PROMPTS.md. Step two is to customise the text content you want generated. Step three is to paste it into nanobanana2 and generate. The recommended settings table gives the model as a specific image preview model, the resolution as 2048 by 1152, which is 2K in 16:9, and the aspect ratio as 16:9. There is also a core capabilities section, listed as the persona the content should be written in. It covers deep searching popular notes on a platform to distil what made them work, summarising and using analogies so a complicated topic is understandable, a visual-first approach with more graphics, a data-driven strategy where each module contains specific numbers, and golden quotes for the key takeaways where needed. Those capabilities are the content instructions that sit alongside the style instructions. A style prompt tells the model how the slide should look, and the capabilities tell it what the slide should say, which is why the workflow has a separate customisation step for your own text. ## It installs as a skill, and the package says so plainly The package manifest is where the shape of the project becomes unambiguous. The name is style-prompt-studio, the version is 1.0.0, and the type field is skill with the main entry point set to skills.md. So this is packaged to be installed as an agent skill rather than run. The usage block gives the install command as npx skills add with the repository URL, and an example of what you would then ask for: generate a PPT slide in retro-pop style about Y Combinator. The files list is the skill manifest itself, with skills.md, README.md, PROMPTS.md, CLAUDE.md, LICENSE and an .openclaw directory. The keywords name ppt, slides, nanobanana2, image-generation, prompts, design, 302.ai and claude-code. The capabilities object is a machine readable version of the same persona list, with content transformation, viral content logic, summarization, visual first, data driven and golden quotes as the keys. That is worth noticing for a different reason: it means the skill description is structured metadata that a tool can read, not only prose in a readme. The top level of the repository matches. There is a PROMPTS.md, a prompts directory, a styles directory, a demos directory holding the example images, a scripts directory, a skill.json, a skills.md, a CLAUDE.md and a .openclaw directory. There is a package.json but the test script is a placeholder that echoes that no tests are specified, which is the expected shape for a repository of text. ## What this repository is not It is worth naming the confusion, because the name slides and the description about generating PPT slides both invite it. There is no renderer here. No code takes your content and produces a .pptx or an .html deck. The deliverable is a prompt, and the generation happens in someone else's image model, at their resolution and with their interpretation of the words. The styling is also fixed. Each prompt names a palette, a background treatment, a typography direction and a layout. You can change the words you want on the slide, but the design is the one the prompt describes, and the 16:9 at the end of every string is not a suggestion to the model so much as a constraint you are trusting it to respect. What you can take from the repository is the prompt engineering itself. The consistency of the structure across the styles, the specificity of the colour values, the repeated insistence on where the title and the key statistics go, and the separate content persona with its requirement that every module contain a concrete number, are all reusable without the specific aesthetics. The maintenance picture is modest. The licence is MIT, the last push was on 2026-04-07, and the repository publishes no GitHub releases, so there is no version series to track beyond the manifest's 1.0.0. ## Conclusion Adopt this repository if you produce presentation slides with an image generation model and want reproducible visual styles instead of describing a look from scratch each time. Do not adopt it if you need editable slide content, because the output is a generated image and the text in the prompt is fixed rather than your own copy. Before you use it, check three things: which model you will paste into, since the recommended settings name a specific image preview model, whether your content fits a single poster-like slide, since the extended styles are described as optimized for that, and that you are comfortable that the prompts specify exact hex values that the model may not follow precisely. ## FAQ ### What is Style Prompt Studio? A collection of English prompt strings, one per visual style, that you paste into an image generation model to produce a 16:9 presentation slide. There are 19 styles in total, and it ships as an installable skill with the package type set to skill and the main entry point at skills.md. ### How do I install and use the slides skill? Run npx skills add with the repository URL, then ask for a slide in a style, for example to generate a PPT slide in retro-pop style about a topic. The quick start itself is described as no code: pick a style from PROMPTS.md, customise the text, and paste it into the image model. ### What are the recommended settings for the slides prompts? The recommended settings table gives the model as a specific image preview model, the resolution as 2048 by 1152, which is 2K in 16:9, and the aspect ratio as 16:9. Every style prompt also ends with 16:9 as a constraint on the output. ### What is the difference between the original styles and the extended styles? The first eleven examples are the original Seedream 4.0 outputs, and the added eight use Nano Banana 2 / Pro, with their current English and Chinese entries temporarily reusing the same image. The extended eight are described as optimized for single-slide, poster-like image generation. ### Does this repository generate PowerPoint files? No. There is no renderer, the deliverable is a prompt, and the generation happens in an external image model. The repository ships a styles directory, a prompts directory, a PROMPTS.md, demos of the example images and a scripts directory, and its test script is a placeholder that reports no tests specified. ## Sources - [AAAAAAAJ/slides on GitHub](https://github.com/AAAAAAAJ/slides) - [Issues](https://github.com/AAAAAAAJ/slides/issues) - [License: MIT](https://github.com/AAAAAAAJ/slides/blob/main/LICENSE) - [README](https://github.com/AAAAAAAJ/slides/blob/main/README.md) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/aaaaaaaj-slides