# Image Generator ## Writing the prompt This is the first item of the main panel. This is where you'll be creating your prompts to generate images. ![Example of a Roman Soldier prompt with the Studio Engine](https://1025298504-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FLY2rXNTpHIzl6JFxbxzb%2Fuploads%2FucAHNfFME83G8i866fmE%2Fqs_3_generate.png?alt=media) If you need help to improve your prompts, we recommend checking out [**How to do better Prompts**](https://krea.gitbook.io/krea-wiki/getting-started/how-to-do-better-prompts)**.** **Pro Tip**: When clicking the ***Generate** button*, you can request up to 4 simultaneous generations at the cue and they will generate even if you switch to another tool or do something else. ## 1. Select an AI Engine ![Select an AI Engine](https://1025298504-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FLY2rXNTpHIzl6JFxbxzb%2Fuploads%2F0xTbkLTlZOQ7GRqxc3ex%2Fqs_4_select_ai_engine.png?alt=media) You can choose a specific AI engine based on your needs: * **Real**: Focused on photorealistic images and high-quality visuals. * **Studio**: Excels in creating dark, colorful images with lots of detail. * **Krea**: Our custom model is trained on 500k trending text-to-image prompts. Similar to Real. * **Luna**: Trained on high-quality dreamy, painting-like, and ethereal images. * **Illustration**: Ideal for creating stylized flat visuals like logos, icons, and SVGs.

Different results on the same prompt depending on the engine

## 2. Advanced Settings In the advanced settings, you can fine-tune your project: * **Dimensions**: Choose the aspect ratio for your project (1:1, 3:4 or 4:3). * **Prompt Influence**: Regulates the trade-off between obeying the prompt and aesthetics. * **Negative Prompt**: Remove elements you don't want in your image. * **Diffusion Steps**: Set the number of steps for the denoising process. * **Setting a Seed**: If you want to reproduce your results, you can fix the seed of the generations. ![Advanced Settings Overview](https://1025298504-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FLY2rXNTpHIzl6JFxbxzb%2Fuploads%2FdZOPbsqbyoWcHiUsNoAd%2Fqs_5_advanced_settings.png?alt=media) ### 2.1 Dimensions

Different image generations (same prompt) based on dimensions

### 2.2 Prompt Influence ![](https://1025298504-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FLY2rXNTpHIzl6JFxbxzb%2Fuploads%2FZbVgMOmiTb9sAUI3Xsf3%2FScreenshot%202023-06-26%20at%2015.27.39.png?alt=media\&token=4c8b5057-2555-4f25-85bb-3d5fc0feceec) This regulates the trade-off between prompt obedience and aesthetics. By default set to 7. * A lower setting (2) will obey the prompt. * A higher setting (14) will prioritize the aesthetic look.

Different image outputs (same prompt) with changed prompt influence values

### 2.3 Negative Prompt ![](https://1025298504-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FLY2rXNTpHIzl6JFxbxzb%2Fuploads%2FkRV2l2uxIMjkHNuM2SoA%2FScreenshot%202023-06-26%20at%2015.31.17.png?alt=media\&token=c0eca53f-f977-47f7-8afd-ddb89bd86d45) It allows you to remove certain undesired elements from the image, such as low-quality, artifacts, or any objects that appear badly generated given your specific prompt.

An example image generation with a negative prompt to remove beans that were appearing around the coffee

### 2.4 Diffusion Steps ![](https://1025298504-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FLY2rXNTpHIzl6JFxbxzb%2Fuploads%2F87LlsHNIQvC4aI4hKh6t%2FScreenshot%202023-06-26%20at%2015.39.41.png?alt=media\&token=eb5ab779-af99-4a3f-a4c8-5819993607f0) Regulates the number of denoising steps. High values will make generations more detailed but it will take more time to compute them.

Impact of modifying the diffusion steps based on the same prompt

### 2.5 Freezing the seed ![](https://1025298504-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FLY2rXNTpHIzl6JFxbxzb%2Fuploads%2FxcpklAFyGEZJ5ujNFQB1%2FScreenshot%202023-06-26%20at%2015.41.37.png?alt=media\&token=d878203f-2991-40c2-aeef-eb6e6d86c76a) Setting any arbitrary number here (like 42 or 12345) will allow you to consistently get the exact same results with the same prompt. There is a creativity trade-off since you will get less variations (since the same prompt will generate the same results) but it allows you to ensure you always get what you are looking for.

For example, you can try to get the same result using: * **Prompt**: A cup of coffee * **Model**: Real * **Guidance Scale**: 7 * **Seed**: 2951064947 And there you have it! You have created your first image and you are now ready to start ideating with KREA.