> For the complete documentation index, see [llms.txt](https://help.craftsmanplus.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.craftsmanplus.com/intelligence/specialties/asset-generation.md). # Asset Generation Asset Generation creates the media your playable needs, right inside the chat. Describe what you want and INTELLIGENCE produces it, ready to place in your project or save for later. You can generate: * **Images:** backgrounds, characters, product shots, icons, and other artwork. * **Vector shapes:** icons, logos, and illustrations as editable SVG shapes. * **3D models:** objects and characters for use in Code Layers. * **Audio:** background music, sound effects, and voiceover. This specialty **produces assets** rather than giving advice. Other specialties lean on it too: when the Builder Agent needs an asset that doesn't exist yet, it can generate one as part of a larger task, such as filling in missing artwork during a reskin. *** ### Working With Generated Assets Generated assets appear in the conversation: images and 3D models in a grid, audio as a player you can listen to right there. Generated assets aren't added to your project or saved automatically. Each result gives you two actions: * **Add Layer:** adds the asset straight to your current scene as a new layer. * **Save to Asset Library:** saves the asset to your Asset Library so you can use it later, without adding it to the project yet. Nothing is saved or placed unless you choose one of these actions, so you can generate freely and only keep what you want. To build on a result, hover it and click the check mark to select it. Selected results attach to your next message, so you can ask for changes or variations. ### Choosing the Right Model INTELLIGENCE has a range of generation models to choose from, each suited to different kinds of work, like illustration, photography, text in images, or icons. It picks the model that best fits your request, so you get a wide variety of art styles without choosing a tool yourself. Describe the look you want in plain words, like *"flat vector style"*, *"photorealistic"*, or *"hand-painted watercolor"*, and INTELLIGENCE chooses accordingly. It starts with the most cost-effective model that fits and saves the higher quality models for when they're needed. *** ## Image Generation Image Generation creates custom images from a text description. Describe what you need and INTELLIGENCE generates it in the conversation, usually in 20 to 60 seconds. ### What You Can Generate Image Generation is open-ended. Common uses include: * **Backgrounds and scenes:** environments, gradients, textures, and backdrops. * **Characters and mascots:** original characters, mascots, and avatars in a style you describe. * **Product imagery:** product shots and hero images, with or without a background. * **Icons and UI elements:** badges, coins, buttons, and decorative accents. > *"Generate a background image of a tropical beach at sunset."* > > *"Create a cartoon mascot holding a coffee cup."* > > *"Make a product shot of a sneaker on a clean white background."* ### Providing Context To get assets that match your style, give INTELLIGENCE reference material along with your prompt: * **Reference images:** attach images so INTELLIGENCE can match a style, color palette, subject, or composition. * **URLs:** share a website or app store listing, and INTELLIGENCE can draw on the visual style and branding it finds there. * **Brand guidelines:** upload a brand guidelines PDF so generated assets follow your approved colors and visual identity. The more reference material you give, the more closely results reflect your style rather than a generic take on your prompt. > *"Here's our website. Generate a hero background that matches this style."* > > *"Using the attached brand guidelines, create a set of icons in our brand colors."* ### Shape and Size Images are generated at the aspect ratio that suits your request, like 9:16 for a portrait background or 1:1 for an icon. Ask for a specific shape, like *"landscape, 16:9"*, if you need one. When you use a generated image in your project, it uses the **optimized version** of the file, the same way uploaded assets do, to keep your file size in check. ### Removing Backgrounds INTELLIGENCE can remove the background from any image: one it generated, one you attach, one in your Asset Library, or one already on a layer. It returns the subject on a transparent background at the original size, without changing the subject itself. When you ask for a new image, you can also ask for a transparent background up front. > *"Generate a red sports car with a transparent background."* > > *"Remove the background from the hero image in scene 1."* *** ## Vector Shapes For icons, logos, and illustrations, INTELLIGENCE can generate vector graphics. Vectors stay sharp at any size and save to your Asset Library as **Shape** assets. Place one on a Shape layer and each color in it gets its own color selector, so you can recolor it to match your brand without generating it again. See [Shapes](/play/layers/shapes.md). For simple geometry, like stars, arrows, and speech bubbles, INTELLIGENCE draws the shape directly instead of generating it, which is faster and keeps it fully editable. > *"Generate a vector icon of a shopping bag in a flat style."* > > *"Create a simple logo badge as a vector."* *** ## 3D Models INTELLIGENCE can generate 3D models from a description or from a reference image. Models take several minutes to generate, and you can turn them to inspect them right in the chat. > *"Generate a low-poly 3D treasure chest."* > > *"Make a 3D model of the sneaker in this photo."* {% hint style="info" %} 3D models work in [Code Layers](/play/layers/code.md) only. Ask INTELLIGENCE to build a feature that uses the model, like a product that spins as the user drags it. {% endhint %} *** ## Audio Generation Sound makes a playable more engaging: background music sets the mood, sound effects reward interaction, and voiceover guides the player. Describe the audio you want and INTELLIGENCE generates it, with a player in the conversation so you can listen right away. Just say what you want and INTELLIGENCE picks the right type. If a request is unclear, it asks a quick question, like whether you want background music or a one-off sound effect. {% hint style="info" %} INTELLIGENCE can't *hear* audio. It works from your prompt and the clip's details, so it can name, place, and save a clip, but it can't judge how it sounds. {% endhint %} ### Music Background tracks from a description of mood or genre. * Describe the **mood or genre** you're after. * Set a **length** from 3 to 60 seconds. The default is 30 seconds. * Ask for an **instrumental** track when you don't want vocals. > *"Generate upbeat background music."* > > *"Make a calm, ambient loop for a puzzle level, instrumental, about 20 seconds."* ### Sound Effects Short sounds for interactions and feedback. * Describe the **sound**: a click, a coin pickup, a whoosh. * Set a **length** from 0.5 to 30 seconds, or let INTELLIGENCE fit it to the sound. * Ask for a **loop** for continuous ambience, like rain or crowd noise. > *"Generate a button click sound effect."* > > *"Create a looping forest ambience with birds."* ### Voiceover Spoken lines generated from text. * Provide the **line** to speak, up to 900 characters. * Describe the **voice** you want, like *"a calm female voice"*, or ask for one of the built-in voices by name, like Adam, Alice, Charlie, or Sarah. * Set the **language**, or let INTELLIGENCE detect it from the text. * Ask for a faster or slower **delivery**. > *"Generate a 'You Win' voiceover line."* > > *"Say 'Tap to play' in an upbeat male voice."* *** ### Good to Know * **Saved assets are reusable.** Anything saved to your Asset Library is available in every project, like any other upload. * **Original work only.** INTELLIGENCE won't generate copyrighted characters or real people, and flags prompts that name a trademarked brand. * **Generation uses AI credits.** Images, audio, and 3D models each count toward your company's AI allowance. * **Animated Banners.** Generated images work in banners. Audio and vector shapes can't be used in banner projects. *** ### Tips * **Be descriptive.** For images, mention the subject, art style, composition, and background. For audio, mention the mood, genre, or exact sound, and how long it should be. * **Name the style you want.** Words like *"pixel art"*, *"flat vector"*, or *"photorealistic"* help INTELLIGENCE choose the right model. * **Ask for a transparent background when you need one.** If an image will sit over your own scene, ask for it up front. * **Use vectors for icons and logos.** They stay sharp at any size and you can recolor them later. * **Say which kind of audio you want.** "Music", "a sound effect", or "a voiceover" up front gets you a result faster. * **Generate, then decide.** Nothing is saved automatically. Generate a few options and keep the ones you like. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://help.craftsmanplus.com/intelligence/specialties/asset-generation.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.