Generate Images

Overview

Trovella's image generation runs a two-pass Gemini pipeline: optional style analysis on your reference images, then the actual generation with your prompt and references woven in. Generated images are stored in your workspace and surfaced in a gallery you can come back to.

Generating an Image

1. Open Images from the sidebar.
2. Write a prompt that describes what you want. Be specific — "a Scandinavian living room in late-afternoon light, oak floors, large window with a fjord view" beats "a nice living room."
3. Pick an aspect ratio (square, portrait, widescreen, banner, and more) and a resolution (auto / 512px / 1K / 2K / 4K).
4. (Optional) Add up to 14 reference images by drag-and-drop. For each one, choose a category:
   • Subject — preserve what's in the photo (the person, object, scene)
   • Style — match the artistic style (brushwork, color palette, medium)
   • Composition — match the framing and layout
5. Click Generate image. The job runs in the background; the right pane shows progress (queued → analyzing style → generating → saving) and the finished image when it's ready.

Reference Images

Reference categories drive how Gemini interprets each image:

• A style reference triggers a separate Gemini call that writes a paragraph describing the style. That description gets woven into the main prompt so the style anchors at the text level — not just as a trailing image.
• Subject and composition references go straight into the generation call with category-specific framing.

Images are resized to 1024px on the longest side before upload, so high-resolution references are fine.

Saving Reference Groups

Working with the same set of reference photos across many generations? Save them as a group:

1. Load the references you want.
2. Click the save icon next to "Reference images."
3. Give the group a name.

To load a saved group later, click the folder icon and pick from the list. You can also update an existing group with your current images, or delete one you no longer need.

Prompt Library

Save prompts you reuse often:

1. Click the bookmark icon next to the prompt label.
2. Give the prompt a short label.
3. Open the Library popover to load, edit, or delete saved prompts.

Gallery

Every image you generate is saved to your gallery automatically. Open the Gallery tab to:

• Browse paginated thumbnails (12 per page).
• Click an image for full detail: prompt used, aspect ratio, resolution, every reference image in the original snapshot, and the full prompt Gemini actually received.
• Download the original file.
• Load settings — copies the prompt, ratio, resolution, and references back into the Generation tab so you can iterate.
• Delete removes the image from your gallery and from object storage.

Reference snapshots are stored alongside each gallery row, so older generations stay reproducible even if you later edit or delete the reference group they came from.

What's Stored

• The generated image lives in private object storage (Google Cloud Storage), scoped to your workspace.
• The gallery row holds your prompt, the structured prompt Gemini received, the style description (if style references were used), and a frozen JSONB copy of every reference image used at generation time.
• Reference groups and saved prompts are scoped to you within your workspace.

Limitations

• Maximum 14 reference images per generation.
• Generation typically takes 10–30 seconds depending on resolution and the number of style references.
• If a generation fails, the gallery row stays as failed with an error message — you can delete it from the gallery view.