Three Ways to Add Images

1. Unsplash Stock Photos (Default)

The AI automatically uses high-quality Unsplash photos. Just describe your content — the AI picks appropriate images. No action needed.

2. AI-Generated Images

Ask the AI to create custom visuals:

"Generate a hero image of a futuristic city at sunset"
"Create an illustration of a team working together"
"Generate a product mockup showing the app on a phone"

Generated images are saved to the Media Library with a "Generated" badge.

3. Upload Your Own

Drag and drop files into the chat or prompt area:

Brand logos, icons
Product photos
Team photos
Wireframes or sketches

Uploaded images have priority over stock photos — the AI uses them first.

File Attachments

The AI can read and analyze what you attach:

PDF — Extract content, tables, data
DOCX — Word documents
XLSX — Spreadsheet data
PPTX — Presentation content
Images — Visual analysis and description
Video — Screen recordings, watched frame by frame (see below)
Archives — ZIP and the like, for bringing an existing project in

Attach by drag and drop, with the paperclip, or by pasting from the clipboard — a pasted screenshot is named with its timestamp. The limit is 50 MB per file and 50 MB per upload.

Attachments go to the model you picked in the chat, so pick one that reads images before you attach a screenshot or a recording — see Models & Thinking.

Video & Screen Recordings

Attach a screen recording and the agent watches it instead of guessing from your description. It works in passes: a coarse one across the whole clip to find the moment that matters, then a close look at the seconds around it, frame by frame. Each frame carries its timestamp drawn into the picture, so the agent always knows which moment it is holding.

Runs of identical frames collapse into a single image marked as unchanged — that the screen did not move is a finding, not a gap.

If the recording has sound, the narration can be transcribed as well. That is usually where the intent lives: "watch — when I click here, the form resets" is in the voice, and no frame shows it. A silent recording is reported as silent.

Every look appears in the chat as its own card, naming the window and the frame rate it used.

Recording a bug report that works:

Keep it short — the seconds around the problem, not the whole session
Say out loud what you expected and what happened instead
Show the whole window if the URL or a console error matters

Media Library

Every project has a Media Library (toolbar icon):

View all project assets
Download any file
See "Generated" badge on AI images
Delete unused assets

Image Analysis

The AI can analyze existing project images:

"Analyze the hero image and describe it"
"What colors are used in the logo?"
"Describe the current homepage layout"

Tips

Upload images before asking the AI to use them
Use descriptive filenames: "hero-banner.jpg" not "IMG_1234.jpg"
For best quality, upload images at least 1200px wide
AI-generated images work best for illustrations and backgrounds