Tools
Image Studio models, sizes, edits
Image Studio makes new pictures from a written prompt, or edits your own photos when you add them as references (pictures the AI copies from). You type what you want, pick a model, a size and a shape, and press Generate — the exact credit price shows on the button before you press it, and the result appears under Latest.
Updated · View as Markdown
Steps
- Open Image Studio in the sidebar (page
/app/image). - Type in the prompt box ("Describe the image you want to create..."). A counter shows how many characters you have used.
- Optional — to edit or match a real product: press Reference photos 0 of N and upload JPG, PNG, WebP or HEIC photos. The number N is how many the chosen model accepts, up to 16. In your prompt, say what to change, for example "put this bottle on a marble table at sunset".
- Press the model chip and pick a model. Use Search models… to find one by name. A dot shows its state: Smooth (running normally), High volume (busy, may take longer) or Paused (can't be picked right now).
- Press Model parameters and choose Size (1K, 2K, 3K or 4K, depending on the model) and Aspect ratio (the shape, for example 1:1, 4:5 or 9:16).
- If the model can make several at once, press the 1x chip and choose up to 4 pictures per press.
- Press Generate · N credits. The picture appears under Latest. Press Download on the tile, or open it full size.
At a glance
| Credits | Shown on the Generate · N credits button; at the time of writing, 2 to 21 credits per picture depending on model and size |
| Max length | Prompt has no fixed cap; a warning appears past a model's own limit, and 100,000 characters is the hard ceiling |
| Sizes | 1K, 2K, 3K, 4K or Standard, depending on the model; shapes from 1:1 to 21:9 |
| Inputs | A prompt; up to 16 reference photos depending on the model (JPG, PNG, WebP, HEIC) |
| Wait | The tile shows "Creating image N" until it is ready |
| Best for | Product photos, ad pictures, picture edits, and starting frames for videos |
Model families
Image Studio lists the models that are on sale right now; the list can change. The main families:
| Family | What the controls allow |
|---|---|
| GPT-Image 2 and GPT-Image 2.5 | GPT-Image-2 Standard takes up to 16 reference photos and goes up to 4K; GPT-Image 2.5 makes 1K pictures |
| Nano Banana, Nano Banana 2, Nano Banana Pro | Up to 14 reference photos; Nano Banana Pro goes up to 4K |
| Seedream 4.0, 4.5, 5.0 | Up to 10 reference photos; the failure tip names Seedream for realistic people and faces |
| Flux Kontext, Flux 2.0 | Flux Kontext takes 1 reference photo — good for editing a single picture with a prompt |
For live per-model credits, see AI video and image model pricing.
Use when / Avoid when
Example
Upload a photo of your perfume bottle under Reference photos. Pick a Nano Banana Pro model, set Size to 2K and Aspect ratio to 4:5, and write: "Image 1 on a wet black stone, soft morning light, water drops, product photo." Press Generate. To reuse a demo instead, scroll to Start from an example and press Recreate — it fills the prompt and model for you.
Troubleshooting
| Message on screen | What to do |
|---|---|
| "This image model declined this request (its content rules are strict about realistic people). Your credits have been returned. Tip: try the Seedream model for realistic people and faces." | Switch to a Seedream model and press Generate again. |
| "This model is temporarily unavailable — please try a different model." | Pick another model. |
| "N of M images failed — your credits were returned. Try again." | Press Generate again. |
| "Image generation failed. Your credits have been returned." | Try again, or change the prompt or model. |
| "Reference photos trimmed to N for [model]." | The new model takes fewer photos; the extra ones were removed. |
| "Price unavailable, try again in a moment" | Wait a moment and press the retry button next to it. |
| "Image Studio is temporarily unavailable." | The page is paused; try again later. |
Saved pictures also appear in Images (page /app/images). They are kept for 30 days, then can be deleted — download what you want to keep. On the free plan (if you have never paid), downloads carry the REVIRAL.AI logo; see the free-plan logo.
FAQ
How much does one picture cost?
It depends on the model and size. The Generate button shows the exact credits for your choice before you press it.
Can I edit my own photo?
Yes. Add it under Reference photos, then describe the change in the prompt, for example put this bottle on a marble table.
What happens if a picture fails?
The failure banner says why, and your credits are returned. If the model refused a realistic person, the message suggests trying the Seedream model.
How long are my pictures kept?
Saved pictures are kept for 30 days, then can be deleted. Download anything you want to keep.
Can I turn a drawing or sketch into a picture?
Yes, with the three GPT-Image-2.5 models: GPT-Image-2.5, GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst. Other models do not show the sketch pad. Press 'Draw a sketch', draw with the pen (6 colours, eraser and undo), then press 'Use this sketch'. Your sketch becomes a reference photo. If the prompt box is empty, a ready-made prompt is filled in that keeps your layout and adds nothing new. In our test, Flare followed the sketch most faithfully, and the plain model was faster.
Can I change one thing in a picture I already made?
Yes, on pictures made with the GPT-Image-2.5 models. Press 'Edit this picture' on the result card. The picture is added as a reference photo, and you type only what should change, for example 'change the cup to blue and keep the rest the same'. Press 'Finish / HD' to send it to GPT-Image-2.5 Sunburst for a sharper final version. The Generate button shows the exact credits before you press it.