# Images

> Make pictures from words and reference pictures, then edit them: describe a change, replace or remove the background, erase, expand, enhance. Prices included.

## What you will make

A still picture, in the shape you choose, and then as many edits of it as you like. Pictures are cheap (a still costs from under 1 credit to about 10, depending on the model) and they are the best way to settle how something should look before you spend credits on video: send the picture to [Studio](https://studio.oneinfer.ai/docs/studio) as the first frame, and the clip begins exactly there.

## A picture from start to finish

Say you need the opening frame for a film about the monsoon.

1. Open **Images** in the menu. The stage says **Describe a still**.
2. Write: "A tea stall at dusk in heavy monsoon rain, warm lamp light on wet stone, shallow depth of field." The box takes light, lens and subject.
3. Check the shape in the **Still** pill (it starts at **16:9**), then look at the button: **Render still · 0.35 credits**. The price depends on the model; the default, **MiniMax Image-01**, is the cheapest in the list.
4. Press **Render still**. The picture appears on the stage and in **Recent stills** on the right.
5. Press **Edit** under it to change it, or **Send to Studio** and **Start** to make it the first frame of a clip.

![Images with the model list open and a price on each model](https://studio.oneinfer.ai/docs/images/images-model.webp "Choosing an image model, demo data")

## What you will see

The stage is on top, the prompt box under it, and **Recent stills** on the right (a strip above the box on a narrow screen). Under the box:

| Control | What it does |
| --- | --- |
| **Add to shot** | Adds pictures to guide the still: your cast, places and props from **Cast & World**, or **From device**, **Paste**, **Camera** and **From Vault**. The counter shows how many the model takes. |
| **Taste** | A ready-made look, added to your words when you render. |
| **Still** | The shape (**Shape**) of the picture. |
| The model name | Which image model makes it, with its price. |
| **Render still** | Makes one picture. It shows the price, and reads **Rendering…** while it works. |

**Polish prompt** rewrites your words into a fuller prompt for the model, shape, pictures and Taste you chose. **Render still** stays off until there is a prompt, and you must be signed in.

![The Images screen before a picture is made, with the prompt box, Taste, shape and model](https://studio.oneinfer.ai/docs/images/images-start.webp "Images, demo data")

Pictures made here are one at a time: there is no setting for how many. Resolution is the model's own and is not a choice.

## Pick the model

Press the model button. Each row shows the model's name, what it is good at, and its price per still. At the time of writing:

| Model | Good for | About |
| --- | --- | --- |
| **MiniMax Image-01** | The cheap default; one reference picture | 0.35 credits |
| **MiniMax H3 Image** | Many references (up to 16) | 2 credits |
| **ByteDance Seedream 5.0 Pro** | Up to 14 references, high quality | 4.5 credits |
| **FLUX 3 Image** | Up to 10 references | 10 credits |
| **FLUX 3 Image Klein** | Low cost, up to 10 references | 1.4 credits |
| **Grok Imagine Image** | Text, or up to 5 references | 2 credits |
| **GPT Image 2.6 Sunburst** and **Flare** | The newest, most detailed and the fast one | about 3 credits, depending on the prompt |
| **MAI-Image-2.7** | Accurate text in the picture, up to 5 references, transparent backgrounds | about 3.5 credits, plus about 0.8 a reference |
| **MAI-Image-2.7 Flash** | The same, faster and cheaper | about 1 credit, plus about 0.26 a reference |

Only models with a published price are listed. Studio chooses **MiniMax Image-01** until you change it. If you add more reference pictures than the model takes, Studio switches to **ByteDance Seedream 5.0 Pro** and tells you.

## Shapes

Press **Still** to open **Shape**: **16:9**, **4:3**, **1:1**, **3:4** and **9:16**, limited to what the model can make. The default is 16:9. A model that does not make the shape you were on moves you to the nearest one and says so. Some FLUX models make **1:1** only. **MAI-Image-2.7** picks the frame itself when it works from reference pictures.

## Transparent background

For product shots and logos, **MAI-Image-2.7**, **MAI-Image-2.7 Flash** and **GPT Image** can make the picture on a transparent background. Press **Still** and choose **Background: Transparent**. It works without reference pictures; with pictures added, the choice is turned off.

## Reference pictures

Add pictures with **Add to shot** to keep a face, a place or a look. How many a model takes: **MiniMax Image-01** 1, **MiniMax H3 Image** 16, **Seedream** models 14, **FLUX 3 Image** 10, **Grok Imagine Image** 5, **MAI-Image-2.7** 5 (PNG or JPEG). Some models take none and work from words alone. You can also type **@** and an actor's name to put your cast in the picture, or paste a picture into the box with Ctrl+V.

## Taste

**Taste** works as it does in [Studio](https://studio.oneinfer.ai/docs/studio#taste): it adds its words when you render and leaves your text as you wrote it. Pictures have four looks of their own: **Luxury editorial**, **Travertine catalog**, **1980s magazine cover** and **Street polaroid**, along with the ones that suit video, such as **Film noir** or **Claymation**. The default is **No taste**.

## Change a picture

Press **Edit** under a picture on the stage, or the small **Edit** tag on a reference picture, or open an older picture from **Recent stills** and press **Edit**. The edit panel replaces the stage. Every result is a new picture saved to your Vault beside the original, and the next edit starts from the result. **Back to the original** returns to where you began, and **Done** closes the panel.

![The edit panel with Remove background, Erase painted area, Replace background, Expand and Enhance, each with a price](https://studio.oneinfer.ai/docs/images/images-edit.webp "Editing a picture, demo data")

The **Quick edits** row has a price on each button:

| Button | What it does | Needs |
| --- | --- | --- |
| **Remove background** | Cuts the subject out on a clear background | The picture |
| **Erase painted area** | Removes whatever you painted over and fills it in | Paint |
| **Replace background** | Keeps the subject and draws a new background from your words | Words |
| **Expand** | Extends the picture to a new shape: 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16 or 16:9 | The picture; words are optional |
| **Enhance** | Sharpens and improves the picture | The picture |

At the time of writing these cost about 2 credits for **Remove background** and about 4.4 credits each for **Erase painted area**, **Expand** and **Enhance**, and **Apply change** about 4.5. **Replace background** is priced like a still from **ByteDance Seedream 5.0 Pro**.

To change something by describing it, type in the box ("make it golden hour", "add a red umbrella") and press **Apply change**, or Ctrl+Enter. If you painted an area first, the change applies only inside it.

To paint: press **Paint an area** and pick a brush (**S**, **M** or **L**). **Clear paint** starts again. Painting narrows an edit to that spot, or marks what **Erase painted area** removes.

> [!NOTE]
> Editing needs you to be signed in. A picture that is not in your Vault is saved there first.

## Results and Recent stills

Under a picture:

- **Download** saves it.
- **Save to vault** keeps it (the button then reads **In your vault**). Pictures that come back inline are saved for you.
- **Edit** opens the panel above.
- **Send to Studio** has **Start**, **Look** and **End**: the picture becomes the first frame, a look reference or the last frame of a clip.
- **Clear render** removes it from the stage and keeps your reference pictures.

**Recent stills** lists this project's latest pictures (up to 40), newest first, each with its model. Press one to show it on the stage with its shape. Only pictures made on this screen appear there, not uploads.

## Questions

### How many pictures does one press make?

One. Press again for another; each press is charged.

### Why did my model change when I added pictures?

The model you chose takes fewer reference pictures than you added, so Studio moved you to **ByteDance Seedream 5.0 Pro**, which takes 14, and told you with a message.

### Can I choose the resolution?

Not on this screen. Each model makes its own size. To sharpen a picture, run **Enhance** on the result.

### Does an edit change my original?

No. Each edit saves a new picture. **Back to the original** gets you back to where you started.
