Start here
Understanding tools in Plnty
What a tool is, how it differs from the model underneath it, and how to pick the one that matches what you are trying to explore.
Every AI generation in Plnty begins with a tool. Before learning any particular one, it helps to understand what a tool is, because it works differently from the model picker and empty prompt box you may be used to. ## What a tool is A tool is a card on the Plnty canvas that performs a specific creative task. It combines inputs, settings, an underlying model and Plnty's own logic into a reusable workflow. You drop your reference onto it, adjust whatever settings matter for this run, and press **Run**; the card keeps that setup, so using the same treatment again later does not mean building it from scratch. <!-- IMAGE SLOT: a tool card on the canvas, annotated to show the input area, the settings, the model pill and the Run button. See writing-docs.md § 7. --> ## How tools differ from models The model is the engine that produces an image, video or mesh. The tool determines how your input is interpreted, what gets asked of the model, how many results return and how those results relate to one another. Several tools can use the same model and still produce very different results because each applies its own creative logic. Choosing a tool is therefore a choice about method, not only about the engine underneath it. :::tip[The model is an ingredient. The tool holds the craft.] This distinction is central to how generation is organized throughout Plnty. ::: ## How to choose a tool Start from what you want to happen to your work, then find the tool that does it. The table below covers the tools people reach for most often when exploring an idea. | Tool | Best for | What it changes | What it tries to preserve | | --- | --- | --- | --- | | **diverge** | Exploring genuinely different directions | Setting, palette, material, scale, lighting and camera | The central subject or idea | | **variations** | Testing alternatives along one specific axis | The axis you name | Everything outside that axis | | **antagonist** | Challenging an assumption in your reference | One underlying assumption | The rest of the composition | | **close-ups** | Finding details inside an existing image | Framing and scale | The original design | | **same-family** | Developing related characters or objects | The individual form | The shared visual world | The right-hand column describes what each tool is designed to hold steady rather than a guarantee. A model can still introduce changes the tool did not ask for. The key difference between **diverge** and **variations** is scope. **diverge** changes several visual decisions at once, while **variations** changes one named axis and keeps everything else as stable as possible. ## Example: exploring directions with diverge **diverge** takes one reference and produces six deliberately different visual directions. It varies elements such as setting, palette, material, scale, lighting and camera while preserving the central idea of your source. Drop an image onto the card and press **Run**. Six results come back, and they are designed to differ from each other in setting, palette and camera rather than to be six near-copies. That spread comes from the tool. Before anything is generated, **diverge** composes six separate briefs and applies rules about how far apart its choices are allowed to land, so the results come from six ideas rather than from six rewordings of one. This behaviour comes from **diverge**, not from the underlying model alone. <!-- IMAGE SLOT: original reference on the left, the diverge card in the middle, six resulting directions on the right. Caption states what changed across the six and what stayed stable. See writing-docs.md § 7. --> ## What you can control Each tool handles some of the decisions that would otherwise need to be specified manually, such as how to structure the request, how far apart to push the results and what to preserve between them. You provide the input and, where supported, your own direction. You can also control, where available, settings such as quality, size and number of outputs. When a tool supports multiple models, the **model** pill lets you change the underlying engine without changing the tool's overall method. The **Run** button shows the credit cost before generation begins. :::note[Continue with another tool] Any result can become the input for another tool. Most Plnty workflows move through several tools as the work develops. ::: ## Where to go next - [How a tool works](/docs/using-plnty/using-a-tool/how-a-tool-works/) covers the mechanics every card shares, from input to result. - [Settings, quality and size](/docs/using-plnty/using-a-tool/settings-quality-and-size/) explains the controls available across tools. - [Images](/docs/using-plnty/what-you-can-make/images/) walks through what you can produce and which tools get you there.Every AI generation in Plnty begins with a tool. Before learning any particular one, it helps to understand what a tool is, because it works differently from the model picker and empty prompt box you may be used to.
What a tool is
Section titled “What a tool is”A tool is a card on the Plnty canvas that performs a specific creative task. It combines inputs, settings, an underlying model and Plnty’s own logic into a reusable workflow. You drop your reference onto it, adjust whatever settings matter for this run, and press Run; the card keeps that setup, so using the same treatment again later does not mean building it from scratch.
How tools differ from models
Section titled “How tools differ from models”The model is the engine that produces an image, video or mesh. The tool determines how your input is interpreted, what gets asked of the model, how many results return and how those results relate to one another. Several tools can use the same model and still produce very different results because each applies its own creative logic. Choosing a tool is therefore a choice about method, not only about the engine underneath it.
How to choose a tool
Section titled “How to choose a tool”Start from what you want to happen to your work, then find the tool that does it. The table below covers the tools people reach for most often when exploring an idea.
| Tool | Best for | What it changes | What it tries to preserve |
|---|---|---|---|
| diverge | Exploring genuinely different directions | Setting, palette, material, scale, lighting and camera | The central subject or idea |
| variations | Testing alternatives along one specific axis | The axis you name | Everything outside that axis |
| antagonist | Challenging an assumption in your reference | One underlying assumption | The rest of the composition |
| close-ups | Finding details inside an existing image | Framing and scale | The original design |
| same-family | Developing related characters or objects | The individual form | The shared visual world |
The right-hand column describes what each tool is designed to hold steady rather than a guarantee. A model can still introduce changes the tool did not ask for.
The key difference between diverge and variations is scope. diverge changes several visual decisions at once, while variations changes one named axis and keeps everything else as stable as possible.
Example: exploring directions with diverge
Section titled “Example: exploring directions with diverge”diverge takes one reference and produces six deliberately different visual directions. It varies elements such as setting, palette, material, scale, lighting and camera while preserving the central idea of your source.
Drop an image onto the card and press Run. Six results come back, and they are designed to differ from each other in setting, palette and camera rather than to be six near-copies. That spread comes from the tool. Before anything is generated, diverge composes six separate briefs and applies rules about how far apart its choices are allowed to land, so the results come from six ideas rather than from six rewordings of one. This behaviour comes from diverge, not from the underlying model alone.
What you can control
Section titled “What you can control”Each tool handles some of the decisions that would otherwise need to be specified manually, such as how to structure the request, how far apart to push the results and what to preserve between them. You provide the input and, where supported, your own direction. You can also control, where available, settings such as quality, size and number of outputs.
When a tool supports multiple models, the model pill lets you change the underlying engine without changing the tool’s overall method. The Run button shows the credit cost before generation begins.
Where to go next
Section titled “Where to go next”- How a tool works covers the mechanics every card shares, from input to result.
- Settings, quality and size explains the controls available across tools.
- Images walks through what you can produce and which tools get you there.