CONTENTLY
HELP

[AI ASSISTANT & VISION]

Generate Images with AI: A New Image or a Fill for Your Layers

Describe a picture and Contently draws it as a new image, or fills every selected layer with its own picture. Gemini and OpenAI make the images.

published by Daniel Schulz updated by

Goal

Get placeholder or final artwork into your design without leaving Figma.

Before you start

  • An AI provider that can generate images: Google Gemini or OpenAI (or a local Stable Diffusion server). Groq cannot, and the panel does not offer it in this mode.
  • Optionally, layers to fill: rectangles, ellipses, frames or any layer that has a fill.

Steps

  1. Open Generate Image Click the sparkle button (AI Text, Image & Copilot) in the sidebar and choose Generate Image next to UX Writing. The prompt box now asks you to describe the image you want to generate.
  2. Describe the picture Write what the image should show, for example 'Pixel monsters on soft tinted squares: a white spirit, a phoenix and a red dragon'. A description is required: without one Contently says 'Please provide a description for the image'. To give a reference, use the + button (Upload image or Screenshot selection); only Google Gemini uses it.
    The AI view with Generate Image selected: a prompt for pixel monsters, a white spirit, a phoenix and a red dragon.
    Generate Image: describe the picture in the prompt box.
  3. Choose where it goes With nothing selected, Contently creates a new 1024 × 1024 image in the middle of your view, selects it and zooms to it. With layers selected, it makes one picture for each layer and uses it as that layer's image fill; the fill replaces whatever fills the layer had.
  4. Send and wait Press Enter or the send button. When the picture is placed, a message says 'Image generated successfully!'. A run counts as one AI action, however many layers it fills.
    After sending, the message Image generated successfully appears above the prompt box.
    A finished run: Image generated successfully!
  5. Name layers to steer the result A selected layer's own name guides its picture: a layer called 'Coffee mug' gets a picture of a coffee mug in the style of your description. Layers with a default name such as 'Frame 12' are not used as a subject. Contently also asks for pictures without text, labels or watermarks.

Tips

  • Describe style and mood once, and let each layer's name give the subject.
  • Only Gemini takes a reference image; OpenAI and local servers ignore it.
  • Fill templates later: generated pictures are ordinary image fills, so Smart Templates can still replace them.

Result

A new image layer, or your selected layers filled with matching pictures.

If something goes wrong

  • 'Please provide a description for the image'.

    Type what the picture should show; an empty description is not sent.

  • 'No API key configured for <provider>'.

    The provider that makes the image has no key. Add one in AI Settings → API Key, or choose another provider.

  • 'Gemini image generation is restricted in your region. Please use OpenAI for image generation.'

    Choose OpenAI for image generation.

  • 'Failed to generate any images for selected elements'.

    No layer got a picture. Layers without a fill are skipped; select layers that can hold one, and try again.

  • A message says today's free AI uses are used up.

    The free plan has 12 AI actions a day, shared by every AI feature. A paid upgrade removes the limit.

Common questions

Can I choose the size or shape of the image?

Not in the plugin: images are square, 1024 × 1024.

Which providers can generate images?

Google Gemini and OpenAI, and a local Stable Diffusion WebUI if one answers at the local base URL. Groq cannot make images.

Where are my prompts kept?

In the prompt history of the AI view, shared by its modes: up to 10, pinned ones included.