Skip to main content

​
Overview

Open Canva exposes a focused set of controls so you can generate quickly without tuning low-level parameters. The app automatically handles backend defaults like inference steps and guidance; you only manage the settings below.

​
Model selection

  • Choose up to four models per request. One image is produced for each selected model.
  • All selected models share the same prompt and aspect ratio.
  • Model availability and size support are validated automatically before a run.
Start with a single model, then add more to compare results side by side within the same batch.

​
Aspect ratio and image size

The aspect ratio picker maps to supported image sizes for the chosen models. Available ratios:
Aspect ratioMapped sizeNotes
1:11024×1024 or 512×512Default square options
4:31024×768Standard landscape
3:4768×1024Standard portrait
16:91024×576Widescreen landscape
9:16576×1024Widescreen portrait
2:3Uses nearest supported sizePortrait-friendly
3:2Uses nearest supported sizeLandscape-friendly
The UI will keep your selection aligned with what the backend accepts—no manual dimension entry required.

​
Reference images

  • Upload up to 5 reference images per session and reuse them across prompts.
  • Attach up to 3 references per generation request (platform limit) to guide results.
  • Uploaded references and recent generations appear in the References panel; you can toggle selections before running a prompt.

Reference Images Guide

Learn how reference limits and reuse work.

​
Prompt enhancement

  • Use the Enhance action to expand or refine your prompt before generating.
  • Enhanced text can be kept in sync with the editor or discarded at any time.
  • Enhancement runs client-side against the prompt you provide; no extra settings are needed.

​
What isn’t configurable

Some parameters are handled automatically by the service and are not exposed in the UI:
  • Inference steps and guidance scale
  • Seed selection
  • Negative prompts or style presets
This keeps the interface streamlined while still enforcing safe defaults and model-specific requirements under the hood.

​
Best practices

Start with defaults: Default settings work well for most use cases. Adjust only when needed.
One setting at a time: Change one setting at a time to understand its effect.
Document successful combinations: Keep notes on settings that work well for specific types of images.
Balance quality and speed: Higher quality settings take longer. Find the right balance for your needs.

​
Settings reference

SettingPurposeNotes
ModelsWhich engines run1–4 models per request; one image per model
Aspect ratioComposition and sizeMapped automatically to supported dimensions
Reference imagesGuide style/contentUp to 3 per run; reuse uploads stored in the session
Prompt enhancementImprove prompt qualityOptional helper before generation

​
Troubleshooting settings

  • Increase inference steps
  • Try a different model
  • Check image size settings
  • Ensure prompt is detailed enough
  • Increase guidance scale
  • Make prompt more specific
  • Try a different model
  • Check if prompt is too complex
  • Reduce inference steps
  • Use smaller image size
  • Check model selection (some are faster)
  • Use a seed value for consistency
  • Increase guidance scale
  • Be more specific in your prompt

Model Selection

Learn which models work best with different settings.

Writing Prompts

Combine effective prompts with the right settings.

Reference Images

Learn about reference image settings.

Troubleshooting

More help with common issues.