Creating Your First Image
Generate your first image in LocalBanana with a gallery prompt or your own text. Check references, model settings, and the result before iterating.
Create Your First Image in LocalBanana
For a first attempt, choose one subject and one clear style. Decide what will make the result useful before you generate: for example, a centered product with room for a heading. You can also start from “Generate Image” in the sidebar if you already have a prompt.
- 1
Find a starting image
Open a gallery image detail and read its prompt. Choose a reference with a composition close to your goal. Browse the gallery.
- 2
Open the image composer
Select “Use prompt” in the detail view. LocalBanana opens the composer on the same page; review the supplied prompt and replace the subject or other details you want to change.
- 3
Check the inputs
Confirm the selected model and aspect ratio in the controls. Gallery reuse does not attach the original reference images. If you need your own face or product in the image, upload its source photo and choose a model with reference-image support.
- 4
Generate
Review the cost shown for the current configuration and complete any sign-in or balance requirements. Start generation and wait for the displayed result or status; completion time varies.
- 5
Inspect and save
Open the result and use the checklist below. If it meets your goal, download the image. Keep the prompt, model, and settings with it so you can understand later variations.
Check the Result Before Calling It Finished
- Subject: Is the intended person or object present? Check faces, hands, product shape, and labels where relevant.
- Composition: Is the crop usable, with the needed space for a heading or other content?
- Text: Read every letter and check line breaks. Correct spelling does not guarantee the requested layout.
- Preservation: If you supplied a source image, compare the details you asked to keep side by side.
If something fails, name that specific issue and change one instruction at a time. For an edit, “Edit Image” in Copilot lets you send the image with a natural-language request. Asking to keep a face, pose, or clothing unchanged is a constraint to verify, not a guarantee of pixel-identical preservation.
When the First Result Misses the Goal
- Prompt reuse checklist — Check missing references, model differences, and settings before adding more prompt text.
- AI image editing prompts, tested — Compare source images with actual edited results and look for unwanted changes.