Tips&Tricks

Tips, Tricks, and Examples on How to Prompt Omega

General Tips

1. Be explicit and clear -- avoid assumptions. Specify layer names, file paths, and parameter values when possible.
2. Ask for ideas first -- before asking Omega to do something complex, ask it for ideas, solutions, or methods. Consider using Tutorial/Didactic mode for guided interactions.
3. Retry if needed -- if it doesn't work the first time, ask again with a revised prompt. It often suffices to ask Omega to "try again" (see the reproducibility analysis from the publication).
4. Give feedback -- don't hesitate to call out mistakes and explain how to revise an approach, code, or widget.
5. Use the code editor -- use Omega's AI-Augmented Code Editor to review, edit, and rerun generated code.

---

Example Prompts

Basic Interaction

• "What is your name?"
• "What tools do you have available?"

Image Processing Widgets

• "Make me a Gaussian blur widget with sigma parameter"
• "Make a widget that applies the transformation: y = x^alpha + y^beta with alpha and beta two parameters"
• "Create a widget to multiply two images"
• "Make an image sharpening filter widget, expose relevant parameters"
• "Write a configurable RGB to grayscale widget, ensure weights sum to 1"
• "Can you create a widget to blend two images?"

Image Operations

• "Open this tiff file in napari: https://people.math.sc.edu/Burkardt/data/tif/at3_1m4_03.tif"
• "Downscale by a factor 3x the image on layer named 'image'"
• "Rename the selected layer to 'downscaled_image'"
• "Upscale image 'downscaled_image' by a factor of 3 using smart interpolation (not nearest neighbor)"
• "Split the two channels of the first layer (first axis) into two separate layers"
• "Switch viewer to 3d mode"

Image Analysis

• "How many channels does the image on layer 0 have?"
• "Open this file in napari: https://uk1s3.embassy.ebi.ac.uk/idr/zarr/v0.4/idr0062A/6001240.zarr"
• "Create a napari widget for a function that takes two image layers and returns a 3D image stack of n images where each 2D image corresponds to a linear blending of the two images between 0 and 1"

Image Search

• "Open a photo of Albert Einstein in napari"
• "Find 5 images of cell mitosis and open them"

More widget examples can be found in this table.

---

Multi-Provider Tips

Omega supports OpenAI, Anthropic (Claude), and Google Gemini models. Here are some tips for working with multiple providers:

• Switching providers: You can switch between providers at any time by selecting a different model in the settings widget and restarting Omega.
• Mix and match: Use a frontier model (e.g., Claude Opus, GPT-5.2) as the main model for complex reasoning, and a fast model (e.g., Gemini Flash, GPT-5 Mini) as the coding model for faster code generation.
• Provider strengths: Different providers may excel at different tasks. Experiment to find what works best for your use case.

---

Vision Tool Examples

When a vision-capable model is configured, you can ask Omega to analyze what it sees in the viewer:

• "What do you see in the current view?"
• "Count the bright spots in this image"
• "Describe the structures visible in this microscopy image"
• "What type of cells are shown in this image?"
• "Is there any artifact or noise in this image?"

---

Denoising Examples

The ImageDenoisingTool (available on Linux and Windows, not on Apple Silicon) can remove noise from images:

• "Denoise this image"
• "Remove noise from the selected layer"
• "Clean up this noisy fluorescence image"

---

Segmentation Examples

Omega supports three segmentation algorithms. You can specify which one to use:

Cellpose

• "Segment the nuclei using Cellpose"
• "Use Cellpose cyto3 model to segment the cells"
• "Segment cells with Cellpose, diameter 30 pixels"

StarDist

• "Segment nuclei using StarDist"
• "Use StarDist versatile_fluo model on this image"

Classic (no dependencies)

• "Segment the image using the classic method"
• "Use classic segmentation with watershed to separate touching cells"
• "Segment the foreground using Otsu thresholding"