Playground Guide

Getting Started

1. Choose a tab: Text to Image, Image to Image, or Upscale.
2. Select a model from the dropdown (models are loaded automatically from the API).
3. Configure your generation parameters.
4. Click the Generate (or Upscale) button.
5. Results appear in the Result section. All generated images are added to the Session Gallery.

Text to Image

Prompt

Describe the image you want. Be specific and descriptive. Include style, lighting, composition, and subject details.

Good prompt: "a photorealistic golden retriever playing in autumn leaves, warm sunlight, bokeh background, 4K"

Bad prompt: "dog"

Negative Prompt

Describe what you want to exclude. Common negative prompts include: "blurry, low quality, distorted, deformed, ugly, bad anatomy."

Model Selection

Different models excel at different styles. See the Model Guide for recommendations.

Sampler

The sampling method affects image quality and style. Euler a is a good default. Try DPM++ 2M Karras for higher quality with fewer steps.

Steps

More steps generally produce better results but take longer. 20-40 steps is a good range. Some models work well with as few as 10-15 steps.

CFG Scale

Controls how closely the AI follows your prompt. Lower values (3-5) are more creative; higher values (8-15) are more literal. Start at 7 and adjust.

Dimensions

Width and height in pixels. Common sizes: 1024x1024 (square), 1024x768 (landscape), 768x1024 (portrait). Must be multiples of 64.

Seed

A specific seed produces reproducible results. Use -1 for random. Note the seed from a generation you like to reproduce similar results.

Batch Size

Generate multiple images at once (1-4). Each image uses the same prompt but may differ slightly due to randomness.

Image to Image

Transform an existing image using a text prompt.

1. Upload a source image by dragging it onto the drop zone or clicking to browse.
2. Enter a prompt describing the desired transformation.
3. Adjust the denoising strength:
   • Low (0.1-0.3): Subtle changes, preserves most of the original.
   • Medium (0.4-0.7): Balanced transformation.
   • High (0.8-1.0): Major changes, original mostly ignored.
4. Click Generate.

Upscale

Increase the resolution of an existing image using AI upscaling.

1. Select an upscaler model from the dropdown.
2. Upload your image.
3. Click Upscale.

Upscaling typically produces images 2x-4x larger than the original with enhanced detail.

Tips for Best Results

• Start with a detailed prompt. The more specific, the better.
• Use negative prompts to remove common artifacts.
• Experiment with different samplers — some work better with specific models.
• If results are too similar to the prompt (rigid), lower the CFG scale.
• If results are too random, increase the CFG scale.
• Use the seed to iterate on a generation you like.
• For consistent sizes, use multiples of 64 for width and height.