# Image generation

Make images from text, edit photos, upscale artwork, and train a model on your own style.

## Models for this work

| Model | Best for | Quality | Speed | From |
|---|---|---|---|---|
| [Stable Diffusion XL](https://gpuvault.io/model/stable-diffusion-xl/) | General purpose image work and anything you will want to fine-tune later | Very good | Fast | $0.45/hr |
| [FLUX.1 dev](https://gpuvault.io/model/flux-1-dev/) | Final images, and anything that needs readable text inside the picture | Excellent | Moderate | $0.89/hr |
| [SDXL Turbo](https://gpuvault.io/model/sdxl-turbo/) | Exploring ideas before committing GPU time to a slower model | Good | Very fast | $0.22/hr |
| [Real-ESRGAN](https://gpuvault.io/model/real-esrgan/) | Rescuing footage and images that are too small or too soft to use | Very good | Fast | $0.45/hr |
| [Segment Anything 2](https://gpuvault.io/model/sam-2/) | Background removal and object masking at volume | Excellent | Fast | $0.45/hr |

## What do you need to do?

- **Text to image** — Describe a picture and get it back. https://gpuvault.io/model/stable-diffusion-xl/
- **Product photography** — Studio shots without a studio. https://gpuvault.io/model/flux-1-dev/
- **Upscale and restore** — Make small or soft images sharp and large. https://gpuvault.io/model/real-esrgan/
- **Background removal** — Cut objects out cleanly, at volume. https://gpuvault.io/model/sam-2/
- **Train on your own style** — Teach a model your look from 20 images. https://gpuvault.io/model/stable-diffusion-xl/
- **Batch from a spreadsheet** — One row per image. Come back to a zip. https://gpuvault.io/model/stable-diffusion-xl/

## Questions people ask

### Do I need to install anything?

No. Everything runs in your browser. You pick a model, we start a machine with that model already loaded, and you get a link to a working app. There is nothing to download, no drivers to update, and nothing left on your computer afterwards. It works the same on a Mac, a Windows laptop, or a Chromebook.

### What if I have never done this before?

The workspace opens with a working example already in it, so your first image is one click away rather than a blank screen. Start by changing the words in the prompt box and pressing generate. Most people are producing usable images within ten minutes. If you would rather not learn the tool at all, use a batch job instead: fill in a spreadsheet of descriptions and we hand back the finished images.

### How much will 500 images cost me?

About $0.50. Stable Diffusion XL makes one image in roughly 8 seconds on an RTX 4090, which costs $0.45 per hour. Five hundred images is about 67 minutes of GPU time. Higher quality models cost more: the same 500 images on FLUX.1 dev is closer to $2.20. Every model page has a calculator that works this out for your exact numbers before you start.

### Can I use the images commercially?

It depends on the model, and we show the license on every model page. Stable Diffusion XL permits commercial use of the images you generate. FLUX.1 dev does not, without a separate license from Black Forest Labs. Segment Anything 2 and Real-ESRGAN both permit commercial use. If your work is paid, filter to the models marked as commercially usable and you will not have to think about it again.

### What happens to my files after?

Anything you save to your workspace folder stays in your account and is there next time you open a session. Everything else, including the machine itself and its disk, is destroyed when the session ends. Batch job inputs and outputs are deleted from our storage 24 hours after you download them. We do not use your images or prompts to train anything.

### What if it does not work?

If a session fails to start or a batch job fails partway, you are not charged for it. Failed jobs are refunded automatically, and the error tells you what went wrong and what to change. If the images simply are not good enough, that is a different problem and usually a model choice: try FLUX.1 dev for detail and readable text, or Stable Diffusion XL with a more specific prompt.

---

*Source: https://gpuvault.io/working-on/image-generation/ — GPUVault, GPU rental by the minute.*