# Generating images (/docs/generating/images)
Image generation is the fastest way to get started in SimpliGen.
## The basic flow [#the-basic-flow]
1. In the prompt builder, choose **Image** as the media type.
2. Open a pack and pick an image **preset** (a style or model).
3. Write your prompt, then generate.
Your result appears in the [Gallery](/docs/generating/queue-gallery) when it finishes.
## Cloud or local [#cloud-or-local]
You can run most image presets two ways:
* **Cloud**: generates on our GPUs using credits, with nothing to install. See [Connecting SimpliGen Cloud](/docs/cloud-credits/connecting).
* **Local**: generates on your own GPU for free after a one-time model download. The first time you use a local preset, SimpliGen downloads its models.
If a preset offers both, you will see a **Generate Cloud** button (with a credit cost sub-label) and a **Prepare Local** button (which becomes **Generate Local** once the model is ready). API-only presets show a single **GENERATE IMAGE** button instead.
## Getting better results [#getting-better-results]
* **Match the preset to the goal.** A photoreal preset and an anime preset want very different prompts.
* **Set the right [resolution and aspect ratio](/docs/generating/resolution-duration)** for where the image will be used.
* **Use a [negative prompt](/docs/generating/negative-prompts)** to remove recurring problems such as extra fingers or watermarks.
* **Make a few at once.** Generating a [batch](/docs/generating/seeds-batches) and picking the best is faster than tweaking one at a time.
* **Add a [LoRA](/docs/models-loras/using-loras)** to push a specific character or style on supported presets.
## Refining [#refining]
Found one you almost like? Reuse its seed (shown on the result) and change a few words, or nudge the [advanced controls](/docs/generating/advanced-controls), to iterate while keeping the overall composition.