Skip to main content

Getting Started with Vision FX

Vision FX uses AI models to transform and edit existing images. This guide will help you get up and running quickly.

What you need before starting​

Before you can transform or edit images, you need to install at least one AI model.

AI models are not included by default. They are downloaded through the Model Hub based on your hardware and preferences.

You also need a source image, such as a photo, illustration, design, render, or other image you want to transform.

Step 1 - Install a model​

Open the Model Hub and install a model:

  1. Click Get Models
  2. Browse available models and select a model. If you're unsure, choose the one tagged Recommended for Most
  3. Choose your Hardware Platform — NVIDIA, Intel, or AMD
  4. Choose your Compute Device — GPU or CPU, when available
  5. Review the variant details, including disk space and memory requirements, then click Install

Once installed, the model variant will be available for selection in Model Settings.

Step 2 - Select your model​

After installing a model, choose how it will run:

  1. Open AI Engine > Settings
  2. In Model Settings, choose your installed model
  3. Select a model variant optimized for your hardware

Depending on the selected model and variant, Vision FX will show the available processing methods, such as CPU or GPU, and check whether your selection is compatible.

Vision FX may also suggest optimizations to improve performance or stability.

If you're unsure, choose a variant that is marked as compatible with your system. You can install additional variants later if needed.

Step 3 - Transform your first image​

  1. Open or add the image you want to transform
  2. Enter a prompt describing what you want Vision FX to create or change
  3. Optional: choose a preset style
  4. Optional: enable features such as Structure Guidance or Style Reference if supported by the selected model
  5. Click Generate

You can continue refining your result by adjusting your prompt, changing settings, choosing a different preset, or generating again.

Understanding the workflow​

Most workflows follow the same general process:

  1. Choose a model
  2. Open a source image
  3. Write your prompt or editing instruction
  4. Adjust settings
  5. Click Generate
  6. Refine and iterate

Different models may expose different settings and capabilities. For example, some models may support features such as negative prompts, guidance controls, Structure Guidance, Style Reference, or instruction-based editing, while others may not.

Generating creates a new result based on your source image, prompt, and current settings. You can continue adjusting settings, trying different prompts or presets, and generating again to refine your result.

If you want to explore different directions while keeping an earlier result, duplicate the image before generating again.

Tips for best results​

  • Start with a clear source image that shows the subject or scene you want to transform.
  • Use simple, direct prompts or editing instructions.
  • Use presets to quickly explore different styles.
  • Use Structure Guidance when you want to preserve composition, pose, or layout.
  • Use Style Reference when you want to guide the visual look using another image.
  • Generate multiple results to compare creative directions.
  • Try different models if results vary. Each model has its own strengths.
  • If generation is slow or fails, try a different model or variant, switch processing method when available, or enable options like reduced GPU memory usage if supported.

Troubleshooting​

App controls are disabled

  • Select a model in Model Settings to enable generation features.

No models available

Generation fails, is slow, or not working as expected

  • Try a different model or variant, or switch processing method when available.
  • Reduce image size in Image Settings.
  • Enable options like reduced GPU memory usage if available.

Results are not what you expected

  • Adjust your prompt or try a preset.
  • Try changing transformation strength or guidance settings if available.
  • Generate multiple results and refine your settings.
  • Try a different model. Each model has its own strengths.

What's next?​