The Image tab handles generating and editing images from a text description, with two modes selectable at the top of the form.
### What you see on screen
1. A **Mode** picker: *Text to image* or *Image to image*. It does not change the request sent: it **filters the list of models** offered.
2. A **Model** picker, with its credit-cost badge and a **More info on this model** link that opens the model's technical page at fal.
3. **Every other field is built from the chosen model.** Change model and the form reconfigures itself: the model definition, server-side, decides which settings exist.
4. A **Generate Images** button.

🔺 Generation is **asynchronous**: the button does not block, a message confirms the generation has started and invites you to find it in the **Generations** tab. Nothing stops you launching several in a row — and each one spends credits.
### Text to image mode
You describe the image you want in the prompt field. The model generates an entirely new image, with no source image. The more precise the prompt (style, colours, composition, mood), the closer the result.

Use cases:
- 🎨 **Game assets**: backgrounds, characters, objects or interface icons matched to your project's universe.
- 🖼️ **Quick variants**: produce several versions of the same asset at once, when the model offers a "number of images" setting.
- 📖 **Narrative illustrations**: visuals to punctuate key moments of an interactive story or a training module.
- 🎭 **Moodboards / visual prototyping**: test artistic directions quickly before settling on one.
### Image to image mode (image editing)
You supply one or more reference images plus a prompt describing the changes. The model builds on your existing images, keeping their structure, style or content.

🔺 **Reference images must already exist in the project files**: the picker opens Celestory's file browser — there is no drag-and-drop from your desktop at this point. Import your image into the project first.

Use cases:
- 🖌️ **Asset reskin**: apply a new style to an existing character (winter version, pixel art) without starting over.
- 🧩 **Style fusion**: supply several source images to combine two aesthetics (if the model accepts multiple references).
- 🔄 **Series consistency**: generate several variants of the same character (poses, expressions, outfits) while keeping the reference images.
- ✏️ **Inpainting** (retouching a specific area without touching the rest): possible **only if the chosen model offers a "mask image" field**. The mask is an **image you supply** (with the area to regenerate marked), prepared in your usual graphics tool and imported into the project — Celestory has no built-in masking brush.
### Settings you will meet most often
They depend on the model, but recur regularly:
- **Prompt**: description of the desired image.
- **Aspect ratio** or **size**: output format (16:9 landscape, 9:16 mobile portrait, 1:1 square…).
- **Resolution**.
- **Number of images**: batch generation.
- **Output format**: PNG (max quality, transparency possible), JPG (light) or WebP (light and supports transparency).
- **Reference image(s)** and **mask image** (see above).
- **Quality**, **background**, **input fidelity**: depending on the model.
### Available Image Models
As with text, the list lives server-side and can change without a Creator update. The model picker in the interface always reflects the current offering, with each model's credit cost and a link to its technical page.

🔺 Don't confuse these with the models of the **AI image request block** usable in-game: that one offers a closed list of just two, cheaper models (see [AI Generation — Player Side](AI%20Generation%20Player%20v2.md)).
---