Skip to main content
Drift can look at an image you provide and use what it sees as context — the environment you want reproduced in simulation, the robot layout in a photo, dimensions labeled on a datasheet page. You attach the image once at the start of a prompt; Drift ingests it, then pulls exact labeled values on demand as it plans and builds.
Vision is on by default. There’s nothing to install — just prepend the image path to your prompt.

​
How to attach

Prepend the absolute path to the image, then write your prompt:
Drift reads the image once, forms an inventory of what’s in it, and keeps that inventory available for the rest of the session. Follow-up prompts can reference the image without re-attaching:

​
Supported formats

.png, .jpg / .jpeg, .gif, .webp.
.avif is not supported. If you have a .avif (Chrome’s default screenshot format on some systems), convert first:

​
Session lifetime

The image stays available across turns for the current session only. Starting a new drift session resets — re-attach if you want the image back in context.