How to Generate AI Images in Figma

VM Studio is a free Figma plugin. You can install it in two ways:

1. From Figma: Open any Figma file, right-click the canvas, go to Plugins > Search plugins, and type "VM Studio". Click the result and hit Install.
2. From the web: Visit the VM Studio plugin page on Figma Community and click Install. Next time you open Figma, it will be available.

Once installed, open the plugin from Plugins > VM Studio in the right-click menu, or pin it to your toolbar for quick access.

VM Studio connects to AI providers through their APIs. You'll need an API key from at least one provider. Each has different models, pricing, and strengths.

After signing up and creating a key, open VM Studio in Figma, click the settings icon (gear), and paste your key into the field for the provider you chose. VM Studio stores your key locally in Figma — it never leaves your machine except to authenticate with the provider's API.

Open the model dropdown at the top of the VM Studio panel. The model you pick determines image quality, speed, and cost. Here's a quick overview:

• FLUX.2 Turbo — Fast (~2.6s), cheap ($0.008), good for quick drafts and iteration
• FLUX.2 Klein 9B — Very fast (~1.5s), cheap ($0.011), strong quality for the speed
• GPT-Image 1.5 — Highest quality, handles text in images well, slower (~40s), higher cost
• Seedream v4.5 — Excellent photorealism (~16s), good for product shots and portraits
• Nano Banana Pro — High quality (~19s), strong at creative and artistic styles
• Z-Image Turbo — Fastest (~1.5s) and cheapest ($0.005), useful for rough concepts

Not sure which model is right for your use case? Check our model comparisons with side-by-side outputs from identical prompts.

The prompt is the text description that tells the AI what to generate. Better prompts produce better images. Here are practical guidelines:

• Be specific — Instead of "a dog", try "a golden retriever sitting in a sunlit meadow, soft bokeh background, natural lighting"
• Mention the style — Add terms like "photograph", "3D render", "watercolor illustration", "flat vector", or "cinematic" to steer the aesthetic
• Describe the composition — Use "close-up portrait", "wide shot", "top-down view", "centered" to control framing
• Include lighting — "soft diffused light", "harsh directional light", "golden hour", "studio lighting" all make a difference

With your prompt written, press ⌘ Enter (Mac) or Ctrl Enter (Windows) to start generation. You can also click the Generate button.

While the image generates, you'll see a progress indicator. Generation takes anywhere from 1.5 seconds (FLUX.2 Klein, Z-Image Turbo) to about 40 seconds (GPT-Image), depending on the model.

Once complete, the image is automatically placed on your canvas:

• If you have something selected — The image appears to the right of your selection, maintaining alignment
• If nothing is selected — The image appears at the center of your current viewport

The generated image is a standard Figma rectangle fill, so you can resize it, add effects, mask it, or use it in components just like any other image.

Image-to-image lets you use an existing image as a starting point and describe changes. This is useful for:

• Changing the background of a product photo
• Altering the color scheme or mood of an image
• Adding or removing elements from a scene
• Converting a rough sketch into a polished render

To use it:

1. Select a frame or image on your canvas
2. Open VM Studio — it will detect your selection automatically
3. Write a prompt describing what you want changed (e.g., "change background to a sunset beach scene")
4. Press ⌘ Enter to generate

The result appears as a new image next to your original, so you never lose your source image.

When you need multiple options to choose from, increase the image count before generating. VM Studio will produce several images from the same prompt, each with different outputs.

The images are placed in a grid layout with smart positioning. VM Studio uses a diagonal expansion algorithm to find non-overlapping positions, so your existing canvas content stays untouched. If you generate again within 30 seconds, subsequent images continue in the same grid.

This is particularly useful for:

• Exploring different visual directions for a hero image
• Generating a set of avatar placeholders
• Creating multiple thumbnail options for A/B testing
• Quickly filling a moodboard with variations