Image Generation

This guide walks you through generating and editing images with your Obvious agent — from logos and illustrations to diagrams, mockups, and photo-realistic content.

Generating an Image

The quickest way: describe what you want.

Generate a minimalist logo for a coffee subscription service called "Morning Ritual" — clean lines, earth tones, no text

Create an illustration of a team collaborating around a whiteboard, flat design style, warm lighting

Generate a photo-realistic image of a modern office space with floor-to-ceiling windows and natural wood furniture

The agent generates the image and saves it to your project. The more specific your description, the better the result. Include details about visual style (illustration, photo-realistic, flat design, 3D render), composition, lighting, colors, and mood.

Choosing the Right Quality

Obvious offers two quality tiers for image generation:

Standard is the default — 4x faster than before and great for photographs, artistic images, quick iterations, and exploring ideas. It handles most use cases well, and you'll get results back in seconds.

Pro produces studio-quality results with advanced text rendering. Use Pro when your image includes text, labels, diagrams, flowcharts, wireframes, infographics, logos, or technical illustrations. The text clarity and fine detail in Pro mode are noticeably better for these use cases.

To request Pro quality:

Generate a flowchart diagram showing the user onboarding process — use Pro quality

Sizes and Dimensions

Images come in five size options:

• 1024×1024 — square, good for logos, icons, and social media posts

• 1152×896 — landscape, good for presentations, headers, and wallpapers

• 896×1152 — portrait, good for mobile content, stories, and full-body illustrations

• 1344×768 — wide (16:9), good for cinematic images, thumbnails, and banners

• 768×1344 — tall (9:16), good for mobile-first content and infographics

Specify the orientation in your request if the default square doesn't fit:

Generate a wide banner image of a mountain landscape at sunset, 16:9 aspect ratio

For higher resolution output, ask for 2K or 4K — available with Pro quality for images that need to stay sharp at large sizes or when fine text must remain crisp at scale.

Editing Existing Images

You can modify images you've already generated. Reference the previous image and describe what you want changed:

Change the background to a gradient from blue to purple

Remove the person on the left side

Make the lighting more dramatic — add warm golden hour tones

Add our company logo in the bottom-right corner

The agent uses the existing image as a starting point and applies your modifications. This is useful for iterating on a design without starting from scratch each time.

Character Consistency

When generating multiple images of the same character or subject, describe physical features precisely and reuse the same descriptors across prompts. "A woman in her 30s with short black hair, olive skin, wearing a navy blazer" produces more consistent results across multiple images than varying the description each time.

Reference previous images when making variations:

Generate the same character from the last image, but now sitting at a desk working on a laptop

Embedding Images

Generated images are saved as files in your project. You can embed them in documents, reference them in conversations, or use them as cover images for documents. The agent can set a generated image as a document's cover:

Generate a cover image for the Q1 report — abstract data visualization in blue and white tones, and set it as the document cover

Tips for Better Results

Lead with the subject. Start your description with the main subject, then add style, lighting, and mood details.

Be specific about style. "Illustration" is vague. "Flat vector illustration with bold outlines and a limited color palette" gives Obvious much more to work with.

Use Pro for anything with text. Standard quality handles photographs and artistic images well, but text rendering is where Pro makes a clear difference.

Iterate. Generate a first version, review it, then refine. "Make it more vibrant" or "zoom out to show more of the scene" — small adjustments often get you to the right result faster than rewriting the entire description.